For conceptual purposes, it is useful to think of IMV exposure as a measurable “dose” characterized by (i) intensity, (ii) duration, and (iii) spectral content over time. Practically, we suggest a two-tier dosimetry approach. First, source-proximal monitoring can quantify vocal-fold/upper-airway vibration using a neck-surface accelerometer (or contact microphone) that has already been validated for ambulatory vocal monitoring (Mehta et al., 2012; Lien and Stepp, 2014; Lei et al., 2022). Second, cranial transmission can be quantified using lightweight tri-axial accelerometers placed at reproducible skull locations (e.g., mastoid/temporal bone and forehead/vertex) to capture band-limited RMS acceleration (a_RMS) and/or a vibration dose value (VDV-style cumulative metric) integrated over time within pre-specified bands (e.g., 20–200 Hz and 200–500 Hz). This enables an explicit dose–response analysis and allows matched-dose experimental controls (e.g., external vibro-stimulation calibrated to the same skull acceleration as self-vocalization). Importantly, aging could reduce this cumulative dose through behavioral and sleep changes even if Aβ production remains stable.

At present, a definitive “threshold” for IMV-supported clearance is unknown, and identifying an optimal window is itself a central test of the hypothesis. A conservative starting point is to define naturalistic IMV ranges using ambulatory dosimetry in healthy adults (e.g., typical daily speaking/singing/humming and stable, non-hypoxic sleep) and treat these as an initial safety-anchored envelope. Early mechanistic studies can then probe dose–response by varying duration and spectral emphasis within this physiological envelope while continuously monitoring symptoms and sleep quality.

A key design principle is to separate mechanical dose from cognitive/affective dose. This can be achieved by calibrating external vibro-stimulation to match measured skull acceleration during self-vocalization, and by using active controls that match attention and arousal without adding vocal vibration (e.g., silent reading, paced breathing, or listening tasks). A dose window would be considered “promising” only if biomarker changes track the measured mechanical dose after accounting for sleep architecture, hypoxia indices, and arousal.

We propose three main routes by which IMVs could support Aβ clearance:

Taken together, these mechanisms suggest that a chronic decline in IMV dose could narrow the clearance bottleneck and tilt the balance toward net Aβ retention, even if production remains stable.

The proposed cascade can be summarized as follows:

This cascade is meant to describe a modulatory axis rather than a primary cause. IMVs are proposed to act alongside genetic risk, vascular and metabolic factors, and other established influences on production and clearance.

Insufficient or fragmented sleep elevates dementia risk, and shorter mid-life sleep duration predicts higher incident dementia (Sabia et al., 2021). Acute sleep deprivation in humans elevates Aβ PET signal in hippocampus and thalamus and increases CSF Aβ species (Shokri-Kojori et al., 2018; Ooms et al., 2014; Lucey et al., 2018). Mechanistic work links consolidated slow-wave sleep to enhanced glymphatic flow and coupled slow-wave, hemodynamic, and CSF oscillations that promote solute clearance (Xie et al., 2013; Fultz et al., 2019).

These findings independently support a robust “sleep → clearance → Aβ” link. Within our framework, reduced sleep duration and altered sleep architecture may also reduce nocturnal exposure to benign cranio-cervical vibrations (a component of the overall IMV dose) and thereby remove a potential low-intensity mechanical input that could modulate transport dynamics. We stress, however, that this is a secondary and testable proposition: IMV-related effects, if present, should be evaluated as additive modulation after accounting for sleep architecture, arousals, and hypoxia-related confounders. This also highlights a key confounding structure: sleep quantity/architecture and sleep-disordered breathing (e.g., arousals, intermittent hypoxia) can influence both clearance physiology and the magnitude/pattern of nocturnal cranio-cervical vibration exposure. Therefore, any evaluation of IMV contributions must explicitly control for polysomnographic sleep metrics and OSA-related indices, and should test whether quantified IMV dose explains additional biomarker variance beyond sleep architecture itself. This discrimination logic is reflected in our proposed study designs using concurrent sleep assessment and cranial/neck accelerometry.

Music-based interventions improve mood, behavior, and sometimes cognition in dementia (Leggieri et al., 2019; Ting et al., 2024). Trials in which participants actively sing often report larger benefits than those emphasizing passive listening or usual care, including gains in memory and quality of life for both patients and caregivers (Särkämö et al., 2013; Särkämö et al., 2016; Bleibel et al., 2023).

These effects are typically interpreted through cognitive, emotional, and social mechanisms. We do not dispute those pathways. We suggest an additional biophysical nuance: active vocalization imposes strong, proximal cranio-facial vibrations that are largely absent during passive listening, potentially providing an IMV “top-up” aligned with clearance-oriented improvements.

Crucially, this mechanical interpretation is testable: future trials can include (i) active vocalization versus (ii) matched non-vocal cognitive engagement and (iii) dose-matched external vibro-stimulation calibrated by skull accelerometry, enabling explicit separation of vibrational dose from social, emotional, and cognitive drivers.

Age-related hearing loss is a robust, dose-dependent risk factor for dementia, with higher levels of loss associated with substantially elevated hazard ratios (Lin et al., 2011; Livingston et al., 2020). Randomized evidence indicates that treating hearing loss can slow cognitive decline in higher-risk older adults (Lin et al., 2023; Liu, 2024).

Mainstream interpretations emphasize increased cognitive load, social isolation, and sensory deprivation. Our hypothesis introduces a testable mechanical co-pathway: individuals with substantial hearing impairment may speak less and vocalize less over time, reducing their accumulated IMV dose and slightly biasing clearance against Aβ. This idea does not conflate Deaf identity with disease risk and does not claim that mechanical factors are primary; it simply adds a modifiable behavioral dimension that may coexist with psychosocial mechanisms.

Upper-airway vibration during sleep (often labeled “snoring” or stertor) has been reported across multiple mammalian species (Lonergan et al., 1998; Toth and Bhargava, 2013; Beck et al., 1995; Chopra et al., 2016). Here we treat it primarily as an exposure source of cranio-cervical vibration, not as evidence of an adaptive trait. We therefore do not argue that snoring is evolutionarily conserved because it improves clearance. Instead, we make a narrower and testable point: when upper-airway vibration occurs without hypoxia or marked sleep fragmentation, it may contribute to the nocturnal IMV exposure spectrum. This exposure can be quantified (e.g., via neck/skull accelerometry and respiratory indices) and evaluated while carefully controlling for obstructive sleep apnea, intermittent hypoxia, arousals, and other confounders that are independently linked to amyloid and cognitive outcomes.

As a first check on physical plausibility, we modeled sound propagation and intracranial acoustic fields using finite element (FEM) simulations in COMSOL Multiphysics (see Supplementary Figures S1, S2). We compared three conditions with calibrated sound pressure levels (65 dB SPL), using different reference points: (i) passive listening to a sound source 1 m away, adjusted to yield 65 dB SPL at the ear; (ii) self-initiated vocalization (“singing”), adjusted to yield 65 dB SPL at 1 m from the mouth; and (iii) closed-mouth humming (same calibration as singing). In the 20–400 Hz range, the average acoustic pressure in the brain during self-vocalization was on the order of 60–70-fold higher (≈36 dB) than during passive listening, indicating much stronger near-field vibrational drive for the same perceived loudness. The ratio peaked around 760 Hz, where self-vocalization produced ~550-fold higher average intracranial pressure (≈55 dB) than passive listening, consistent with resonant properties of the vocal tract, although such peaks are likely damped in vivo by tissue absorption and other losses. At higher frequencies the relative advantage of self-vocalization declined as standing-wave patterns changed. These simplified simulations do not capture the full complexity of skull and tissue mechanics, but they support the basic claim that self-generated vocal sounds couple substantially more mechanical energy into the brain than equally loud external sounds heard at a distance.

We emphasize that the FEM results are intended as an order-of-magnitude plausibility and relative-coupling comparison, not a definitive prediction of in vivo intracranial pressure fields. The current model does not fully capture tissue heterogeneity, fluid–structure coupling at CSF interfaces, frequency-dependent damping in living tissues, or the potential influence of vascular pulsatility, all of which may alter attenuation and spatial distribution. Accordingly, a key next step is empirical validation using (i) standardized skull-surface accelerometry during controlled self-vocalization versus passive listening, and (ii) cranial phantoms with tissue-mimicking materials and a fluid layer approximating CSF to calibrate damping parameters and guide subsequent model refinement.

A simple prediction might be that vocally intensive professions (teachers, singers, clergy, broadcasters) would exhibit lower AD incidence or later onset. To date, however, there are no large, well-controlled cohort studies that specifically quantify “vocal load” and relate it to dementia risk. Occupational analyses typically emphasize education, cognitive complexity, and socioeconomic factors (Hyun et al., 2022), and “musician” categories often include both singers and instrumentalists (Balbag et al., 2014; Arafa et al., 2022). Moreover, many voice-heavy jobs involve chronic noise exposure and associated hearing loss—risk factors in their own right (Cantuaria et al., 2024; Cloeren et al., 2025; Zhang et al., 2025).

We therefore view the absence of a clear occupational signal as a reflection of measurement and confounding limitations, not strong disconfirmation. The correct test is not crude job category, but direct, wearable measurement of vocal and vibration dose, with careful adjustment for hearing loss, sleep/OSA, education, and vascular/metabolic risk.

AD cohorts have higher rates of obstructive sleep apnea (OSA), and OSA is associated with increased amyloid and tau burden, adverse CSF profiles, and elevated dementia risk (Emamian et al., 2016; Maimon and Hanly, 2010; Bubu et al., 2020). Snoring intensity correlates with OSA severity. At first glance, this seems to contradict the idea that upper-airway vibrations might be beneficial.

The IMV framework draws a distinction between:

We anticipate that in OSA, any potential benefit from vibration is outweighed by the harms of hypoxia and fragmentation. Consistent with this, treating OSA with CPAP improves slow-wave activity, sleep consolidation, cognition, and several biomarker measures, even though mechanical vibration may be reduced. A central, testable prediction is that after controlling for oxygen desaturation and arousal indices, vibration dose from non-hypoxic snoring (or from safe daytime vocalization or vibro-stimulation) will correlate with better clearance proxies, whereas vibration delivered in the context of OSA will not.

A practical near-term test would be a within-subject crossover study in middle-aged adults:

Primary outcomes could include MRI-based proxies of perivascular and glymphatic flow and short-term plasma or CSF Aβ dynamics. Sleep architecture and cardiovascular parameters would be monitored and controlled. The specific prediction is that Condition B will show more favorable changes in clearance-linked metrics than Condition A, in the absence of differences in cognitive or social engagement.

In individuals with subjective cognitive decline, mild cognitive impairment, or early AD, randomized trials could compare:

An important translational question is how IMV modulation interacts with Aβ-targeting therapies. For example, does increasing IMV exposure before or during treatment with monoclonal antibodies enhance plaque removal, reduce ARIA risk at a given dose, or allow effective treatment at lower doses?

In AD model mice, mechanistic experiments could address:

Prospective cohorts equipped with wearable voice and vibration sensors, detailed sleep and OSA assessments, audiometry, and longitudinal cognitive testing could directly examine whether natural variation in IMV dose predicts amyloid PET signal and dementia risk. In addition to voice-activity metrics, cohorts could incorporate direct cranial vibration dosimetry (e.g., tri-axial accelerometry at standardized skull sites such as mastoid/temporal region and forehead/vertex, alongside a neck-surface sensor) to quantify transmitted IMV dose in frequency bands and cumulative metrics. Rigorous adjustment for confounding factors (sleep quality, OSA, hearing status, education, and vascular/metabolic risk) would be essential. Such data would address limitations of previous occupational and lifestyle proxies.

This hypothesis paper proposes that age-related reductions in low-intensity mechanical vibrations of the head and neck—arising primarily from day-to-day vocalization and benign upper-airway motion—may contribute to Aβ clearance failure in mid-life. We suggest that these intrinsic mechanical vibrations can, in principle, enhance interstitial mixing, perivascular and glymphatic transport, and encounter rates at BBB and meningeal-lymphatic interfaces. In a setting where Aβ production is relatively stable, a gradual loss of this mechanical drive could be one of several factors that make it easier for Aβ to accumulate and form plaques.

The proposal is deliberately modest. It does not claim that IMVs are a primary cause of AD, nor that mechanical effects replace genetics, inflammation, vascular injury, or tau pathology. Instead, IMVs are treated as a potential modifier of clearance efficiency that operates alongside these established mechanisms. The value of the framework is that it connects everyday behaviors (how much people speak or sing, how long and how well they sleep) to biophysical processes (fluid flow and solute transport) that are already known to matter in AD, and it generates a set of concrete experiments that can be performed with current methods.

Within amyloid-centered frameworks, the central question is not whether Aβ is important, but why and when normal homeostatic control is lost. Recent emphasis on clearance has drawn attention to glymphatic and meningeal-lymphatic flow, arterial pulsatility, sleep architecture, and receptor-mediated transport at the BBB. IMVs fit naturally into this clearance-centric view as one more, often overlooked, source of low-frequency mechanical input.

For current therapies that target aggregated Aβ with monoclonal antibodies, the hypothesis has two potential implications. First, chronic differences in IMV exposure—shaped by sleep habits, vocal behavior, and cranio-cervical biomechanics—might contribute to variability in treatment response, for example by influencing how readily the brain can handle mobilized Aβ. Second, if IMV augmentation proves feasible and safe, it could act as a non-pharmacologic adjunct that modestly increases clearance capacity, particularly in preclinical or prodromal stages.

At the same time, it is important to recognize how early the field is in this area. The evidence assembled here is indirect: sleep physiology, 40-Hz stimulation, ultrasound, music and singing interventions, hearing-loss epidemiology, and evolutionary arguments can all be interpreted through other, more established mechanisms. No trial has yet tested whether IMV-focused interventions alter Aβ biomarkers or clinical trajectories. Any discussion of clinical use must therefore remain conditional until such studies are done.

The IMV–clearance hypothesis leads to several specific, testable predictions:

If these patterns are not observed in well-designed studies, the hypothesis should be revised or discarded. The framework is intended to be straightforward to test and, if necessary, to refute.

Several limitations deserve emphasis. First, almost all of the supporting evidence is indirect. The literatures on sleep and clearance, 40-Hz stimulation, ultrasound, music interventions, hearing loss, and snoring each have their own, well-grounded mechanistic explanations. The IMV perspective is an additional layer, not a replacement, and it should be weighed accordingly.

Second, there is a risk of over-interpreting observational associations. The link between hearing loss and dementia, for example, is already well explained by cognitive, social, and sensory factors; any mechanical pathway via reduced self-vocalization is speculative at present. Similarly, the prevalence of snoring-like sounds in mammals does not by itself establish a beneficial role; many evolutionary scenarios are possible.

Third, some aspects of the topic are socially and clinically sensitive. Discussions involving Deaf communities, hearing impairment, and snoring must avoid implying causality or stigma that is not supported by data. The present framework does not claim that Deaf identity or cultural practices cause dementia, nor does it recommend snoring as a health strategy. It distinguishes between modest-intensity vibrations, which might be delivered via vocalization or devices, and pathological patterns such as OSA, where hypoxia and sleep fragmentation are clearly harmful. Future empirical work should be designed with input from affected communities and with careful ethical framing.

Finally, any effects of IMVs are likely to be modest at the individual level. IMVs are proposed as one factor among many—genetic, vascular, metabolic, inflammatory, and lifestyle-related—that shape overall risk. Even if the hypothesis is supported, meaningful public-health impact would probably require combining IMV-related strategies with other established interventions rather than relying on them alone.

Despite these caveats, the hypothesis points to several translational directions that can be explored in a controlled manner. Behavioral programs that emphasize structured self-vocalization—such as humming or singing practice—could be evaluated in randomized trials for their effects on clearance-linked biomarkers, cognition, and quality of life in at-risk or early-stage populations. These interventions would need carefully designed control conditions to separate mechanical contributions from cognitive and social ones.

In parallel, device-based approaches using oral or bone-conduction vibro-stimulation tuned to speech-like frequencies could offer a controllable way to deliver IMV-like input, particularly for people with vocal limitations. Early-phase trials should focus on safety, tolerability, and mechanistic endpoints (perivascular/glymphatic imaging, sleep architecture, autonomic responses) before any claims about clinical efficacy are made. At present, there is no basis to recommend unsupervised vibration regimens as preventive or therapeutic measures.

With respect to Aβ-targeting therapies, validated IMV-based strategies might eventually play three roles: (i) primary prevention by slightly prolonging the period during which endogenous clearance keeps pace with production; (ii) adjunctive support to increase the efficiency of pharmacologic plaque removal or to reduce required drug doses; and (iii) rehabilitative support in later stages, integrated with music-based and cognitive interventions that already have some evidence in dementia care. These possibilities are speculative, but they provide a concrete agenda for translational research.

The next step is to move from conceptual synthesis to direct tests. Short-term human experiments that manipulate IMV exposure under controlled conditions, mechanistic animal studies that examine necessity and sufficiency, and cohort designs that directly measure vocal and vibration dose are all feasible with current methods. More detailed computational models of solute transport that incorporate realistic cranio-cervical mechanics could refine predictions and identify the ranges of vibration amplitude, frequency, and duration where meaningful effects are plausible.

More broadly, the IMV framework encourages a more systematic treatment of mechanical forces in models of brain homeostasis and neurodegeneration. Even if IMVs themselves prove to have only a minor role, clarifying their impact will help refine our understanding of the production–clearance balance at the heart of the amyloid debate and may influence how emerging therapies are combined and timed.