Sound Therapy for Dementia Patients — Calming Sounds, the 40 Hz Question, and What Caregivers Should Know

“Sound therapy for dementia” is not one thing. It is three, with three different bodies of evidence, and most of the confusion comes from letting them blur into each other.

The first is music. This is the old one and the well-studied one: singing groups, a music therapist working one-to-one, or simply a playlist of the songs a person loved at twenty, played on a speaker or headphones. When a care home or a charity says “music helps people with dementia”, this is what they mean, and it is the only one of the three with a body of randomised trials behind it.

The second is steady broadband noise — pink noise, brown noise, a fan, rain on a roof. This is what the player at the top of the page makes. It does one thing, reliably: it covers other sounds. A corridor, a television in the next room, traffic, the clatter of a kitchen. It has no music in it, no memory attached to it, and no evidence that it does anything for dementia. What it can do is make a room quieter in the way that matters, which is a smaller claim and a truer one.

The third is 40 Hz stimulation. This is the newest and the most talked about. It comes from a series of experiments at MIT in which mice bred to develop Alzheimer’s-like changes were exposed to light flickering, and later sound clicking, forty times a second. The mouse results were striking. The human results so far are small, early and mixed, and we go through every one of them below. It is not calming, it is not meant to be, and it is not something to play to someone in the evening to help them settle.

If you only have a minute: music you choose for the person is the one worth your effort; steady noise is a practical tool for a noisy room; and 40 Hz is a research story worth knowing about and not yet a thing to do.

The best single summary is the Cochrane review by van der Steen and colleagues. Cochrane reviews gather every randomised trial on a question, grade each one, and pool the results. The 2018 version included 22 trials with 1,097 people, all living in care homes or hospitals, each receiving at least five sessions of a music-based intervention — singing, playing, listening, or a mix — compared with usual care or another activity. Outcome by outcome:

• Depressive symptoms Moderate-quality evidence that music-based interventions reduce them at the end of the sessions.
• Overall behavioural problems Moderate-quality evidence of a reduction, again at the end of the sessions.
• Agitation and aggression No reduction. Fourteen trials, 626 people, and the pooled effect sat on zero; the review’s own wording is that the interventions “do not decrease agitation or aggression”.
• Anxiety Low-quality evidence of a reduction.
• Emotional well-being and quality of life Low-quality evidence of an improvement.
• Cognition Low-quality evidence of little or no effect. Music does not restore memory or thinking.
• Lasting effects Measured four or more weeks after the sessions ended, the evidence was low or very low quality and pointed to little or no effect.

The review was updated in 2025 with 30 trials and 1,720 people, and the picture held. Against usual care, music probably improves depressive symptoms slightly and may improve overall behavioural problems; it likely does not improve agitation or aggression; and for anxiety, well-being, social behaviour and cognition the evidence was too weak to say. No serious adverse events were reported in any trial. That last line is the practical one: the risk of trying is low, the plausible benefit is mood, and the thing families most want it to fix — agitation — is the thing the pooled trials did not show it fixing.

Personalised music is a special case, and the one with the longest history. Not “relaxing music” chosen by someone else, but the specific songs that mattered to this person, found by asking them and their family. Linda Gerdner tested it in a crossover trial published in 2000: 39 people with severe cognitive impairment in six Iowa care homes heard either their own preferred music or generic classical “relaxation” music for 30 minutes twice a week, six weeks of each. Agitation fell significantly more during and after the individualised music than during the classical music. It is small and old, but it is where the personal-playlist approach comes from.

The largest programme built on that idea is Music & Memory, which trains care-home staff to make personal playlists and hand residents a music player. Its evaluations disagree. Thomas and colleagues (2017) compared 98 US nursing homes that adopted it in 2013 with 98 matched homes that did not, using routine assessment data on close to 13,000 residents with dementia in each group: the proportion of residents coming off antipsychotic and anti-anxiety medication rose in the participating homes while staying flat or falling in the comparison homes, and reported behavioural problems fell more; depressed mood did not differ. It is a before-and-after comparison, not a randomised trial. Kwak and colleagues (2020) then ran a randomised crossover trial of the same programme in 10 nursing homes with 59 residents over 14 weeks and found no statistically significant difference in agitation, behavioural symptoms or psychotropic medication use — trends favoured the music, none reached significance, and the authors questioned whether busy staff can deliver it as designed. McCreedy and colleagues (2019), in a four-home pilot with 45 residents with advanced dementia, added a detail that explains some of the disagreement: agitation measured by direct observation dropped markedly while the person was actually listening, staff-rated agitation fell over six months, but the administrative records that large studies rely on showed no change at all. The effect, in other words, may be real, mostly confined to the minutes the music is playing, and invisible to the paperwork.

One-to-one music therapy by a trained therapist is a different intervention again. Ridder and colleagues (2013) randomised 42 care-home residents with moderate to severe dementia to six weeks of individual music therapy and six weeks of standard care, in either order. The disruptiveness of agitation rose during standard care and fell during music therapy, a medium-sized difference, and psychotropic medication was increased more often during the standard-care weeks. It is exploratory and the authors ask for a larger trial.

The honest summary for a caregiver: music the person recognises and loved is worth trying — for mood, for engagement, and for the minutes it is playing. Do not expect it to abolish agitation on the day’s worst afternoon; the pooled trials say it will not. Do expect that trying costs little and harms nobody, and that the person’s response — a foot tapping, a word of a lyric, a face that softens — tells you more than any study can.

This is where the searches are most desperate and the evidence is thinnest, so it is worth being exact about what is known.

Sleep problems are common in dementia: more waking in the night, lighter sleep, and a day–night rhythm that drifts. They wear caregivers down, and the Cochrane review of non-drug approaches to them — Wilfling and colleagues, 2023 — notes that they are among the reasons people move into residential care. That review found 19 randomised trials with 1,335 people. They tested light therapy, daytime physical or social activity, carer training, daytime sleep restriction, back massage, electrical stimulation, and combinations of these. Not one of the single interventions tested was music or any other sound, which is the first thing to know: there is no body of randomised evidence that any sound improves sleep in people with dementia. The review’s overall conclusion was that no single or combined approach could be identified as suitable for widespread use; daytime physical and social activity may slightly lengthen night-time sleep and carer-focused interventions may help modestly, all at low certainty.

The music review above did not measure sleep as an outcome at all. So when a page tells you that a particular playlist or frequency is “proven” to help people with dementia sleep, it is either extrapolating from studies of people without dementia, or making it up. What is reasonable to say is narrower: a steady, quiet, unchanging sound in the bedroom can cover the corridor noises and household sounds that wake a light sleeper, and a familiar piece of music can be part of a predictable bedtime routine. Those are plausible, low-risk and unproven.

Sundowning is the name for restlessness, confusion or agitation that appears or worsens in the late afternoon and early evening. Everyone in dementia care recognises it, yet, as a 2016 review by Canevelli and colleagues points out, it has no agreed definition, no standard assessment tool, no reliable estimate of how common it is, and at the time of that review no randomised trial had specifically tested any approach to it, drug or otherwise. A 2019 scoping review reached the same place: the most accepted explanation is a disturbed body clock, and trials to guide management are scant. So there is no evidence-based “sound for sundowning”. What there is, is the experience of caregivers that a calmer, quieter, more predictable late afternoon — curtains drawn before it gets dark, the television off, one familiar sound instead of several competing ones — is easier to get through than a loud one. Sound can be part of that. It is not a fix for it.

The one place sound and sleep in dementia have been measured together is the 40 Hz research, as a side finding in a 22-person trial described below. It belongs in that discussion, not in advice about bedtime.

The 40 Hz story is real science, and the consumer version of it has run a long way ahead of the data. Here is each study, who or what was studied, and what it found. The 40 Hz gamma page goes deeper into the mechanism and the company-run trials; this is just the part a caregiver needs.

• Iaccarino et al., 2016 · Nature · mice The foundational paper. Mice engineered to develop Alzheimer’s-like pathology were exposed to light flickering at 40 Hz. Amyloid levels fell in the visual cortex and the brain’s immune cells changed their behaviour; other frequencies did not do this. Light only, mice only.
• Martorell et al., 2019 · Cell · mice The sound paper. Seven days of 40 Hz auditory tones improved spatial and recognition memory in the same kind of mice and reduced amyloid in auditory cortex and hippocampus; adding light spread the effect further. This is the study that put “40 Hz sound” into app stores. It is a mouse study.
• Clements-Cortes et al., 2016 · Journal of Alzheimer’s Disease · 18 people An exploratory pilot that predates the sound work in mice. Eighteen people with mild, moderate or severe Alzheimer’s received 40 Hz sound and vibration twice a week for six weeks, compared with watching DVDs. Scores on a short mental-status test rose across the 40 Hz sessions and not across the DVD sessions. Six people per severity group, no blinding described, and the authors themselves call the results promising and in need of further research.
• He et al., 2021 · Alzheimer’s & Dementia: TRCI · 10 people A feasibility trial. Ten people with mild cognitive impairment due to Alzheimer’s used an audiovisual 40 Hz device at home for an hour a day for four or eight weeks. It was safe, tolerated and adhered to; their brain activity entrained to the stimulus; imaging and spinal-fluid proteins showed preliminary changes. It measured whether people could do it, not whether it helped them.
• Cimenser et al., 2021 · Frontiers in Systems Neuroscience · 22 people The sleep and daily-living trial. Twenty-two people with mild to moderate Alzheimer’s were randomised, 14 to an hour a day of synchronised 40 Hz light and sound for six months and 8 to a sham. The active group had fewer active periods at night on a wrist sensor, and held steady on a daily-living scale while the sham group declined. Fourteen against eight is a very small comparison; the authors describe it as justifying longer trials.
• Chan et al., 2022 · PLoS One · 15 people Feasibility and pilot studies from the MIT group. Fifteen people with mild probable Alzheimer’s used a 40 Hz light-and-sound device daily for three months, blinded and randomised against a control. The active group showed less enlargement of the brain’s ventricles and less shrinkage of the hippocampus on MRI, better scores on one memory test, and steadier daily rhythms. The authors say this supports a larger, pivotal trial. They do not claim more than that.

Read those six together and three things are clear. First, the strong results are in mice. Mice engineered to overproduce amyloid are a model of one feature of Alzheimer’s, not the disease as it happens to a person over decades. Second, every human study is small — ten, fifteen, eighteen, twenty-two people — and each was designed to check safety and feasibility and to look for hints, not to prove benefit. Third, none of them used a phone speaker. They used purpose-built devices delivering precisely timed light and sound, usually together, with eyes closed, for an hour every day for months, with brain recordings to confirm the stimulus was actually driving a 40 Hz response. A 40 Hz tone from an app is the same frequency and not the same intervention.

So, plainly: no human study shows that 40 Hz sound slows or improves dementia in people, and none of the researchers involved claims that it does. Larger trials are running; the honest position is to wait for them. The demonstration button at the top of the page plays a 40 Hz isochronic pulse through speakers — a curiosity for the caregiver, not a routine for the person.

If you decide to try sound, the details decide whether it helps or becomes one more thing in the room.

When in the day

Before the difficult time, not during it. If late afternoon is the hard stretch, start something calm and familiar an hour before it usually begins, while the person is still settled. Music with meaning is for the daytime, when the person can engage with it; steady noise is for when the job is covering other sounds — a busy corridor, a rest after lunch, the hour before sleep. Do not play music at bedtime that will make them want to get up and dance, and do not play anything at all through the whole night.

How loud

Quieter than you think. The test is whether you could hold a normal conversation over it without raising your voice; if not, it is too loud. A person who cannot tell you a sound is uncomfortable will show you instead, and by then it has been uncomfortable for a while. Start from silence and raise the volume slowly.

For how long

Short sessions, then reassess. Ten minutes is plenty for a first try; twenty is the limit here. The care-home trials mostly used sessions of around half an hour a few times a week. Nothing in the evidence supports leaving a sound running for hours, and a sound that is on all the time stops being a cue and becomes wallpaper.

What to watch

The person, not the clock. Good signs are a softening face, a slower breath, humming, a foot moving in time, eye contact, a name or a memory surfacing. Warning signs are turning the head away, covering an ear, frowning, pulling at clothing, getting up, or any increase in restlessness. Some people with dementia become more agitated with sound, not less — it is one more thing to process — and the only way to find out which kind of day it is, is to watch.

How to stop

Fade rather than cut. A sudden silence can be as startling as a sudden sound. Turn the volume down over ten or twenty seconds, then stop. If the person is distressed, stop straight away and do not try again that day; nothing is gained by “finishing the session”.

Hearing aids and cochlear implants

Hearing loss is common in older age and common in people with dementia. The 2020 Lancet Commission on dementia — Livingston and colleagues — lists hearing impairment among the potentially modifiable risk factors for dementia and reports that using hearing aids appears to reduce the excess risk. In practice: check the hearing aids are in, switched on and have working batteries before you judge whether any sound is doing anything. A hearing aid also changes the sound — it can make broadband noise harsh or music tinny — so try the sound with it in and with it out, and trust the person’s reaction over your own ears. Cochlear implants process sound very differently again, and steady noise in particular may be unpleasant through one; ask the implant clinic rather than experimenting.

The television and the competing-noise problem

The commonest mistake in care settings is adding a sound to a room that already has three. A television in the corner, a radio at the nurses’ station, a trolley in the corridor and then, on top, “calming music” is not calm; it is a fourth thing. If you are going to use sound, first remove sound. Turn the television off, not down. Close the door. Then, into a quieter room, add one sound, and the person has something to attend to instead of something to filter.

Care-home realities

Staff are stretched, devices go missing, headphones get tangled, and a playlist made for one resident ends up playing to a lounge full of people who did not choose it. The randomised trial of Music & Memory that found no significant effect raised exactly this: whether staff can deliver personalised music as intended. What helps: a small speaker per room rather than headphones that need fitting; a written note of the person’s few favourite pieces where a new carer can find it; sessions attached to an existing routine, after breakfast or before the evening meal, rather than to the day’s worst moment; and one named person who checks the hearing aids. A family member visiting once a week can make the playlist.

Sound is a comfort measure. The things that change the course of a bad week are clinical, and they are worth asking about directly.

• “Could this agitation be pain?” People with dementia often cannot say where it hurts. Ask whether a pain assessment has been done and whether regular pain relief is worth trying.
• “Has anything changed medically?” A urinary infection, constipation, dehydration, a chest infection or a new medicine can each show up as sudden confusion or agitation. A change that comes on over hours or days is a medical question first.
• “When was the medication last reviewed?” Sedatives and antipsychotics carry real risks in dementia and are meant to be reviewed regularly. Ask what each medicine is for and whether it is still needed.
• “Is there a music therapist, and how do we get a referral?” If the answer is yes, that is the version of sound with trial evidence behind it.
• “Can a personal playlist go into the care plan?” Written down, with the person’s own music and a named time of day.
• “Has hearing been checked?” And are the hearing aids working.

Then, honestly, what this site offers. The calm-room player at the top of this page makes pink or brown noise on the device, at a low starting volume, for ten or twenty minutes, and stops by itself; nothing is stored or sent anywhere. The noise generator is the same engine with five colours — white, pink, brown, green and violet — and longer timers, and the pink noise and brown noise pages explain each one. The main generator plays binaural beats, isochronic pulses and pure tones over a noise bed, twenty minutes at a time. Binaural beats need headphones to work at all, which is one reason we do not suggest them for someone who cannot tell you the headphones hurt; isochronic pulses and noise do not. The mobile app is the one for longer and background sessions: 44 presets, 34 ambient sounds across nature, noise, meditation, water, fire, music and environment categories, and a Frequency Lab for building your own. None of it slows dementia. The seniors page and the sleep guide cover these tools in older age more generally, and the 40 Hz gamma page is where to go for the whole research story rather than the caregiver’s summary above.