In this review paper, we highlight the importance of recognizing the effects of normal ageing on hearing aid fitting because it has a considerable impact on an individual’s understanding of speech and ability to benefit from hearing aids. Auditory processing and cognition decline throughout early, middle and late adult life and undermine our ability to hear in all types of listening situations. “Normal ageing” refers to non-pathological changes in cognition and auditory ability with age (1). We do not discuss pathological changes with age such as mild cognitive impairment or dementia. Whilst considerable research efforts continue to be deployed to understand the association between hearing loss and dementia, we believe it is particularly important to address normally-ageing older adults because they represent the vast majority of patients attending audiology clinics. In the UK National Health Service (NHS), older adults (over 55 years of age) represent 78% of adult audiology attendances and 87% of those fitted with hearing aids (2); and reported hearing problems appear more strongly associated with age than the degree of peripheral hearing loss, which suggests that processes other than peripheral hearing play a role (2, 3). For the purposes of this review we define “peripheral” changes as those restricted to the outer, middle and inner ear and auditory nerve. We define “central” changes as those within the central auditory nervous system, from the level of the cochlear nucleus up to the cortex, which includes elements of cognition. The aim of this review is to determine the relevant factors in auditory processing and cognition that might have implications for the way in which we set hearing aids, and to define the best approach to setting hearing aids for normally-ageing older adults based on the evidence in the literature. Whilst we do not discuss the relationship between hearing and dementia, it is likely that any recommendations regarding setting hearing aids for those with reduced degrees of cognition due to normal ageing will be particularly applicable to those with pathological cognitive decline.

We first provide a brief review of age-related change to central auditory processing and how this alters dependence on different elements of speech (Section 2). We then offer a short review of the basic concepts of cognitive change with age (Section 3), and how a decline in some elements of cognition also alters speech perception, particularly undermining listening in more challenging circumstances (Section 4). This understanding then enables us to determine how hearing aid parameters and fitting procedures might affect older adults’ listening ability. In Section 5, we derive a set of evidence-based principles for fitting hearing aids to older adults, and assess specific hearing aid parameters against these principles. We review evidence for an association between some elements of cognition and the benefit, or increased impairment, that may be caused by hearing aid processing strategies. We particularly focus on evidence that addresses the benefit, or harm, caused by fast or slow-acting compression speeds for those with different degrees of cognition. We also assess why the evidence may not always appear consistent. We then offer a discussion of how one might determine practical fitting guidelines for audiologists who predominantly see older adults in clinic (Section 6). Finally, we discuss gaps in the current evidence-base and what further research might be needed (Section 7).

The central auditory nervous system (referred to simply as the “auditory system” in subsequent text) is a uniquely complex sensory network, engaging peripheral sensory organs, multiple processing layers in the brainstem, the auditory cortex and wider cortex, culminating in conscious perception and understanding, as well as extensive “top-down” control. The processing of different aspects of speech is distributed across the auditory system. For example, some neurons are tuned to overall amplitude modulation from lower to higher levels of the system (4) and others are phase-locked to periodic patterns in the acoustic detail of speech (5, 6). Segregating and grouping sounds depends on multiple binaural processes from the cochlear nucleus up to the cortex, dependent on effective neural synchrony (7). Multiple pathways in the cortex, largely between temporal and frontal lobes, support a hierarchical process of word recognition, integration into phrases and sentences, syntax (grammar) and semantics (meaning) (8–10). Higher-level processes also depend on the relevance of sounds and lower-level processes can be enhanced via top-down control, enabling selective processing (11).

Age-related changes to the underlying neural infrastructure cause progressive “central” auditory deficits with increasing age (1, 12–17). As “lower” and “higher” level processes are interrelated, the effects of peripheral and central decline can be difficult to dissociate. Age-related cochlear damage generates a degraded input to the auditory system. The degraded nature of the input is exacerbated by the loss of synapses and auditory nerve fibers, often referred to as “hidden hearing loss.” This synaptic loss particularly affects fibers that are sensitive to louder sounds and may particularly impair speech-in-noise (SIN) perception (18–20). With increasing age there is a general reduction in the density of connections in the brainstem and cortical structures involved in auditory processing (14, 21). A deterioration in brainstem structures and the cortex impairs timing information, gap detection, localization, spectral processing and efferent control of the hair cells (22–25). These combined changes can undermine auditory scene analysis, impair an individual’s ability to attend to a target speaker (26) and reduce the ability to distinguish elements of speech in quiet (SIQ), non-familiar accents and SIN (27, 28).

Temporal information in speech cues can be differentiated across a number of frequency bands (29): information about the speech envelope (ENV) is carried primarily in the frequency band below 50Hz; information about voicing and periodicity primarily in the frequency band between 50 and 500Hz; and information about temporal fine structure (TFS) in the frequency band above 500Hz. ENV and TFS are the most important cues forming the basis of speech perception, frequency perception and localization (30). However, for those with sufficient hearing, ENV and TFS provide redundant cues that support processing of the speech signal (31). A difference in the pitch of voicing helps differentiate individuals in multi-talker situations (32). Vocoder experiments that replaced TFS with a carrier signal or noise, which was modulated by ENV, demonstrated that only a small number of frequency bands are required for successful speech perception in quiet even when TFS was lacking (33). However, the same was not true when speech was presented in a noisy background. Adding back TFS to a signal progressively improved SIN performance (34, 35). The ability to make use of ENV with little TFS can also be evidenced by the success of cochlear implants, which provide a grossly reduced representation of spectral detail (29, 36). Besides supporting speech perception in noise (36–38), TFS also supports binaural localization and separation of competing sounds (39), perception of pitch and music (29, 30) and perception of motion (40). Individuals with peripheral hearing impairment cannot make good use of TFS which can lead to difficulties hearing in noisy situations (41).

Comparisons between young and old listeners with matched normal hearing, intelligence and education suggest that speech perception declines with age, demonstrating a greater effect than can be predicted by hearing thresholds (3, 42–44). Monaural TFS sensitivity declines constantly throughout early-to-late adulthood (45), whereas monaural processing of ENV is less affected (46, 47). Binaural processing of both TFS and ENV declines with age, although declining TFS sensitivity has some association with the degree of hearing loss whereas the association between ENV sensitivity and hearing loss is weak (46, 47). Besides TFS, cognitive performance has also been found to predict speech perception. Although both TFS sensitivity and some cognitive domains decline with age, they do not appear to be directly related to each other once the effect of age is removed (48). TFS sensitivity was found to be a stronger factor than cognitive ability and ENV sensitivity in predicting speech perception in quiet for normal hearing listeners, and cognition the stronger factor for predicting performance in noise (3). Reduced SIN perception with increasing age can be more generally associated with reduced cognitive abilities (49), reduced coding accuracy and TFS sensitivity (50, 51), poorer binaural processing (52) and sound segregation (26). Sensitivity to periodicity perception also declines with age, affecting the perception of intonation (53), discrimination of voicing (54) and the emotional content of speech (55, 56). Some elements of speech perception are therefore specific to auditory processing, whereas others are dependent on domains of cognition not specific to auditory processing.

“Cognition” refers to the processes of acquiring, retaining and using knowledge (57) and has many constituent elements, including reasoning, memory, speed, knowledge, reading, writing, maths, sensory, and motor abilities (58). One can differentiate cognition from auditory processing because the latter is specific to the analysis of sound, whereas the former is not. Whilst the auditory system processes sound, cognitive processes develop meaning and understanding from the sound. However, it should be recognized that this is a simplistic description because of the inter-related nature of the systems which together deploy “bottom-up” and “top-down” processes. Some elements of cognition decline with age, but some remain stable or improve (59). Cognitive abilities can be separated into two basic domains (60, 61): “fluid intelligence,” characterized as the speed and ability to resolve problems in novel situations; and “crystallized intelligence” that is an accumulation and use of skills, knowledge and experience. Fluid intelligence generally peaks at about 20 years of age and declines at a consistent rate throughout adult life, whereas crystallized intelligence increases with age. As a result, an individual’s overall performance in psychometric tests of intelligence remains largely stable through most of adult life (60, 62). It should be noted that the range of ability between individuals also increases with age (21, 60, 63) so age alone is not a good measure of specific elements of cognitive function. Individuals of different ages may therefore perform equally well at a complex task, but may be using a different blend of cognitive abilities to complete the task based on the strength of different cognitive domains (64). Accordingly, an overall measure of “general cognition” is not a useful concept when considering a specific complex task, such as listening. Cognitive reserve, developed through education and other lifetime experiences, appears to provide some individuals with greater resilience against cognitive decline (65). This, and other genetic, environmental, health and lifestyle factors, will mediate any cognitive change with age (63) and an individual might further modify function via cognitive training and physical exercise (66). In short, it is important to determine the specific elements of cognition that relate to different tasks rather than make any assumption of homogeneity amongst an age group.

Several theories have been proposed to describe the process of decline in some cognitive abilities. The processing-speed theory (67) suggests that a decline in speed results in processes failing to complete in time to be useful, and that slower processes reduce the amount of available processing capacity. A reduction in processing capacity, caused either by a reduction in processing speed or by other processes of decline, could undermine episodic memory (recollection of personal experiences) and the initiation of more complex processes (68). Note that episodic memory declines more with age compared to semantic memory (recollection of facts) and procedural memory (unconscious performance of tasks) (59). A second theory suggests that there is a limit on the amount of processing resources (69). Based on this idea, Baddeley & Hitch (70) introduced the concept of “working memory” (WM) as a system that stores and processes information relevant to a current task. A third theory postulates that an age-related failure to filter out irrelevant information is the cause for WM to become “cluttered,” thus reducing available capacity (71). Each of the individual theories cannot explain all of the features of cognitive decline and it is likely that there are multiple processes with cumulative effects (59).

One way to categorize the basic concepts of fluid and crystallized intelligence into further subtypes of cognition, particularly useful in relating specific processes to speech perception, is provided by the Cattell-Horn-Carroll model (72, 73) (Figure 1). The model describes a three-level hierarchy of cognitive abilities: general ability, broad domains and narrow domains. Broad domains are basic characteristics that manage a range of behaviors. Narrow domains are highly specialized and may be specific to certain types of task. Auditory processing (Ga) is included as a broad domain in the CHC model (Figure 1). Some elements of overall sound processing and perception are specific to Ga, whereas some elements rely on other narrow domains. Note that the generalized concept of fluid intelligence is not specifically shown in the model, but includes most of the broad domains of the model, i.e., those other than crystallized intelligence (Gc). Nevertheless, in general terms, abilities in the wide domain of fluid intelligence are those that tend to decline with increasing age.

Note that ongoing refinements to the model introduce further broad and narrow domains beyond those shown here (72). Episodic memory is contained within the definition of learning/encoding efficiency (Gl) in the CHC model (Figure 1) (73). Executive function is not commonly included in the Cattell-Horn-Carroll model and there is no clear consensus on its definition (73). It may be thought of as a cognitive “control” function that mediates narrow domain abilities, although its three core elements may be considered as separate domains that: enable “shifting” between tasks; monitoring tasks and “updating” WM as necessary; and “inhibiting” other tasks or automated responses in order to complete a task (74).

In this section we focus on the elements of cognition that may underlie impaired listening ability. Not enough is currently known about how specific narrow domains of cognitive ability (Figure 1) relate to auditory performance in specific listening situations. However we do know that WM often plays a role and is frequently associated with SIN performance (75, 76), understanding fast speech (77) or with general auditory performance across a range of tasks (78, 79). WM is more predictive of SIN performance in older age groups compared to younger groups (80) or in more challenging conditions, such as a lower signal-to-noise ratio (81). Older adults with greater WM resources may be better able to adapt to difficult listening situations (78, 82). WM is often used as a broad term and encompasses both storage and processing-dependent WM tasks. These different types of tasks can have different associations to different types of listening situations (75, 83–85). Executive function is another cognitive factor that is often associated with listening and the degree of listening effort (86). Inhibition, which a sub-domain of executive function (Figure 1), and processing speed are also cognitive functions associated with reduced auditory performance (87–89). Different domains of cognition appear to be more strongly predictive of listening ability for more difficult listening tasks (90, 91). A meta-analysis (92) provided an overview over various cognitive functions and their association with different SIN tasks. It showed that SIN performance correlated with measures of processing speed (r = 0.39), inhibition (r = 0.34), WM (r = 0.28), and episodic memory (r = 0.26). These are all narrow cognitive domains that deteriorate throughout adulthood (62, 67, 68, 71, 93) and their effects may be, in part, additive.

One consequence of the changes to hearing and cognition may be an increase in effort when listening (94, 95). Listening effort is defined in the Framework for Understanding Effortful Listening (FUEL) (96) as an allocation of cognitive resources to overcome listening obstacles. It is a finite resource and its capacity will be expended at different rates dependent on the inherent difficulty of a task and an individual’s motivation to overcome obstacles in listening. Listening effort likely engages multiple neurological systems (97) and is an important factor in speech perception. It should be considered in the context of age-related cognitive decline because it may also affect the motivation of individuals to comply with treatment or engage socially (98, 99). In addition, hearing aids have the potential to reduce, or increase, listening effort (100–104).

The relative importance of hearing loss, auditory processing and cognitive function for predicting overall auditory performance will depend on the specific listening situations encountered (76), which will vary in real-life (105). Different forms of SIN test engage different narrow cognitive domains (92). SIN tasks can be distinguished according to the target and masker signal, and both of these affect the type of processing needed for successful listening (92). The type of masking noise has a considerable effect; individuals perform better in steady-state noise compared to multi-talker babble and the latter yields stronger associations with cognitive function (85, 92, 106–108). An intelligible masking signal provides both energetic and informational masking, and is most likely to divide the attention of a listener, offering the greatest cognitive challenge (75), although other cues such as the different gender or fundamental frequency of the signal and noise source can overcome some effects (109).

A number of speech models have attempted to conceptualize the role of different cognitive functions within the pathway of speech understanding. Gordon-Salant et al. (21) adapted a theoretical model (110) to describe a bottom-up process of hierarchical signal processing. The “sensory system” undertakes spectral and temporal processing to discriminate between sound sources. This is further refined by the “perceptual system,” employing inhibition to direct attention, and the “cognitive system” that analyses and identifies words, and develops meaning, feeding back to lower levels of the system, requiring processing speed, WM capacity and semantic knowledge, also engaging long-term memory (Glr). The model by Bronkhorst (11) is broadly similar in principle and highlights the role of attention control, an element of executive function, in which attention is triggered by certain signal characteristics, that then engenders selective processing at lower levels, or “pre-attentive” stages, of the auditory system. The Ease of Language Understanding (ELU) model (111, 112) largely focusses on situations in which there is conflict between an input signal and lexical information stored in memory, e.g., due to an unusual accent, which then requires engagement of a feedback loop to resolve it, particularly requiring WM, executive function and learning/encoding efficiency. Taken as a whole, these models highlight the critical dependence of listening on certain specific narrow cognitive domains. However, given that research studies cannot assess all real-life listening situations, one cannot conclude that these are the only cognitive domains of relevance, but that they represent those most commonly measured in the context of the limited listening tasks employed in studies, and there may well be other confounding factors such as alternative cognitive strategies that individuals use to compensate (64). The models also highlight that poor fidelity of the input signal (via hearing loss or inappropriate hearing aid processing) can cause greater conflict in resolving speech and, ultimately, a failure to do so. Overall, the models support the finding that speech intelligibility is associated with processing speed, inhibition, WM and long-term storage and retrieval (92) and will consequently decline with age as these narrow cognitive domains deteriorate. It should also be pointed out that a decline in some cognitive domains will not only undermine speech intelligibility, but may also impair the perception of speech quality (113), even where intelligibility is unaffected (114), and undermine an individual’s perception of aided speech.

In summary, certain narrow cognitive domains (Figure 1) will be associated with different types of auditory task. A wide array of cognitive tasks are employed across different studies (92), so the presence and strength of associations between cognition and auditory performance will depend on the narrow cognition domains assessed and the task used to assess them, as well as the type of auditory task, its level of difficulty, the type of stimulus, noise and other cues used (64, 115), the degree of context, vocabulary and visual cues (116–118). Furthermore, there is an association between hearing loss and cognitive decline (119) that may act as a confound in studies. Sensory impairment can affect an individual’s performance on cognitive tests, often requiring recall of spoken words or text-based tests, dependent on hearing and vision, so that it is possible that this causes falsely enhanced associations (61, 120, 121). It may therefore be considered unsurprising that the outcomes of research studies are not wholly consistent because they do not employ consistent paradigms.

The preceding sections of this paper have reviewed the literature to determine how age-related changes to cognitive domains and auditory processing alter speech perception and, in particular, an individual’s relative dependence on different speech cues that might be disrupted by different approaches to hearing aid processing. In this section, we aim to apply the evidence to the clinical hearing aid fitting process in order to provide pragmatic guidance to clinicians.

As part of a patient-centered care model (219) the audiologist’s primary role is to offer support and provide benefit to each individual, where the total benefit can be considered as a combination of objective and perceived benefit, preference, and ability to comply with treatment. This perspective means that the “optimal hearing aid fitting” must therefore address all of these aspects of care. In consequence, audiologists require a broad skill set that includes counselling (219) and some level of technical understanding. However, some relevant technical factors, not least compression speed, are rarely provided by hearing aid manufacturers in specification sheets, nor is it easy in our experience for audiologists to acquire such information. This is a fundamental concern because compression speed and compression ratio are primary determinants for the amount of distortion introduced to ENV. Given that audiologists are, in part, dealing with the hearing aid as a “black box,” and that research findings are not always consistent, it is perhaps unsurprising that there are no clear technical guidelines for fitting hearing aids to older adults (220).

However, the research is broadly consistent in a number of ways. First, older adults are more likely dependent on ENV for understanding speech. FAC causes distortion to ENV that can undermine speech perception, whilst emphasizing elements of TFS that cannot be utilized by many older hearing aid users, and might disrupt binaural inputs. Second, hearing aid users tend to prefer SAC on average, although individuals have different preferences. FAC is likely better in quiet for those with good cognitive processes related to hearing, but likely degrades speech intelligibility and the acceptability of hearing aids for those with lower degrees of cognition. Third, the variation in cognitive processes between individuals increases with age, so we cannot know which specific settings offer the most benefit for an individual, nor whether objective benefit will be aligned with their preference.

The combination of peripheral and central auditory decline therefore means that older adults represent some of the most complex patients, as well as the most numerous. Consequently, age-related hearing loss should never be treated as “standard care” simply because of its high prevalence. Determining the correct hearing aid strategy for an individual hearing aid user is therefore a major challenge. One might think that it would be useful to undertake tests of cognition or TFS sensitivity in clinic. However, it is uncertain that a test score could usefully indicate a specific course of action, nor which cognitive tests should be used. Verbally-delivered cognitive tests are affected by hearing loss and there is no well-established test that has been shown to be unaffected by sensory impairment (61, 120, 221). It is unclear whether audiologists are able to conduct cognitive tests sufficiently well and their utility in clinical situations needs evidencing (222). Speech testing might give some indication of overall ability and it has been suggested that it be more widely used (223) because standard audiometry provides no useful evidence regarding auditory processing and cognition. However, it is equally unlikely that a speech test score could equate directly to a specific hearing aid setting and user preference. It is also unclear whether clinical speech-testing is sufficiently sensitive to determine any benefit derived from alternative hearing aid settings. Consequently, in addition to the current lack of applicable tools, it is debatable whether the additional loading on clinical time would yield sufficient benefit to be justified. In any case, many audiology services face demand and resource pressures, particularly in public sector systems with universal treatment, so extending the test battery significantly may not be implementable.

We therefore propose a pragmatic approach to hearing aid setting that could be implemented within typical current fitting appointments. The principles above can be employed to ensure that relevant factors are considered for older adults, which we define nominally as those over 55 years of age in line with the English NHS definition of age-related hearing loss. We believe the following statements concerning hearing aid settings for older adults follow directly from the evidence:

One should recognize that any recommendation regarding hearing aid parameters cannot be prescriptive given the current state of knowledge and considering the wide variability in individual needs. This implies that there must be some process of ensuring that hearing aid settings are optimal, or at least acceptable, for each individual. Overall, we believe that a full consideration of hearing aid parameters needs to be combined with the over-riding principles of patient centered care. This engenders a need to balance the key elements of the hearing aid fitting process, including verification, validation and counselling. “Verification” is the process of matching hearing aid gain to a prescribed target using real ear measurements (REMs). “Validation” should aim to provide evidence that all hearing aid settings are suitable for an individual. This may include simple tests of loudness discomfort and speech intelligibility, more advanced techniques such as speech mapping, paired comparison approaches (226), or questionnaires and self-reporting tools (227). Finally, “counselling” addresses non-audiometric factors such as the patient’s expectations, self-image and self-efficacy (228). This approach will likely improve the rate of treatment compliance (229, 230). The following statements are not novel, but we believe they are worth re-stating in light of the principles discussed above for setting hearing aids for older adults:

Research studies use a wide array of both cognitive and listening tasks, so the associations observed between them are variable. Although WM tasks predominate in many studies, it is not the only domain of cognition that is relevant (92) and we are yet to fully understand which elements of cognition might relate to every listening condition (64). It would seem prudent to consolidate a range of cognitive tests in research studies. There is also a lack of validated cognitive tests for those with a sensory impairment, because many tasks are delivered verbally or visually, so hearing loss may be a confounding factor (61, 120, 121) and must be considered as part of research design. In large part, studies relate cognitive status to performance in listening tasks, although some studies relate it to user preference. It should be clear that individual preferences are not always aligned to objective performance and may be based more on listening comfort, effort or some other perception of quality, so it would seem sensible to measure both objective performance and subjective benefit in research studies.

Whilst there is variation in the evidence, it appears to us that the balance of evidence suggests that slow-acting compression may be the best default position for most older adults along with other cautious settings for compression ratios and noise reduction, although accepting there will be individual variations. However, further research should seek to confirm the size of the effect. Furthermore, most research on the effect of hearing aid settings considers one parameter in isolation whilst other parameters are held constant. This is unlikely to reflect the overall function of commercial hearing aids and assess the interaction between different hearing aid features. Recent work has demonstrated the interactions between compression speed and other settings, such as noise reduction and directionality (181, 210), so further research should establish whether combinations of hearing aid processing strategies negate the overall effect of individual settings in the user’s perception. It is also important to be able to evaluate balanced measures of speech intelligibility and quality, and relate them to the overall perception of distortion. This indicates a need to further develop clinical tools, such as HASQI and HASPI (148, 149), to determine an objective measure of distortion appropriate for individuals with varying degrees of hearing loss and cognitive ability, that also aligns with user perception. This further begs the question of whether some test of cognitive ability is useful or implementable within clinical practice.

In this paper we have proposed an approach to setting hearing aids for older adults. Whilst this may be derived from the balance of evidence, it is not specifically validated. We have therefore designed a randomized control trial based on the principles discussed in this paper, pre-registered on the Center for Open Science’s Open Science Framework (OSF), at https://osf.io/fdzeh. The SHAOA (Setting Hearing Aids for Older Adults) study has been approved by the UK Health Research Authority and National Health Service (NHS) Research Ethics Committee (IRAS ID 313159), and it has been adopted onto the NHS National Institute for Health Research (NIHR) portfolio. Finally, in suggesting this approach, we make an implicit assumption that audiologists may not consider all of the factors that have been highlighted here, which we have not evidenced. Accordingly, we will also complete an online survey of UK clinicians to evaluate this, under University of Manchester ethics approval.

RW conducted an initial long-form literature review which was reviewed with comments and changes from HD and AH. RW wrote the draft of this paper with comments and revisions from HD and AH. RW is Principal Investigator for ongoing research with HD and AH as academic supervisors. All authors contributed to the article and approved the submitted version.

RW is funded by the NHS National School of Healthcare Science (NSHCS) as part of the NHS Higher Scientific Specialist Training (HSST) scheme, supporting work at the Royal Berkshire NHS Foundation Trust. HD and AH are supported by the NHS National Institute for Health Research (NIHR) Manchester Biomedical Research Centre.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.