Dementia Choirs and Singing Groups: What They Offer and What Has Been Measured

A dementia choir is an open ended singing group for people living with dementia and the people who care for them, run in almost every case by a community musician or a volunteer rather than by a registered therapist, with no assessment, no goals written down and no discharge. I run two of them. They are among the most enjoyable and least studied things happening in dementia support anywhere in the country, and the gap between how they are described in the press and what is actually known about them is wider than for almost anything else on this site.

The person who taught me what these groups are for first came because her husband had a diagnosis and she had been told singing might help him. He came for about fourteen months. After he died she stopped for three weeks, then came back, and she has been in the second row ever since, some five years later. For a long time I filed that under kindness, ours and hers. It is not that. She was never only an escort, the hour had been hers as much as his the whole time, and I had spent years describing my own group inaccurately to funders by putting the person with dementia at the centre of every sentence. Half of the room, most weeks, is there for something the funding application has no box for.

Where this page makes a claim about effect, it carries a label: Tier 1 supported, Tier 2 promising but limited, Tier 3 practice and anecdote. Almost everything here is Tier 3, and Tier 3 covers two very different situations which this page always distinguishes. How the labels are arrived at is on reading arts in health research.

What a dementia choir actually is

It is a sociable, ongoing singing group, open to people with a dementia diagnosis and to whoever comes with them, and its defining structural features are that nobody is assessed on the way in and nobody is discharged on the way out.

Those two absences shape everything else. With no assessment there is no threshold: somebody can arrive with a diagnosis three days old, or five years in and no longer speaking, and both belong in the room. With no discharge the group has no course length, no endpoint and no follow up, which is a large part of why the research literature struggles with it. And because the carer normally attends as a singer rather than a chaperone, the group holds two populations at once, with different needs, in the same forty five minutes.

The person leading it is typically a community musician, a choir leader, an activities coordinator or a trained volunteer, which is not a registered arts therapist, and the difference is legal rather than a matter of standing: music therapist is one of four protected arts therapy titles in the UK, and the HCPC register listed 6,103 arts therapists across all four in July 2026¹. I am a community arts practitioner and I do not use the word therapy for anything I do. The full boundary is on music therapy against community music and on the music therapy pillar.

Groups vary enormously. Some are twelve people around a piano, some are eighty with a conductor and a concert in June, some meet inside a care home and are an activity session with better repertoire. What they share is that they are not treatment, they are not arranged through a clinical pathway, and they exist because somebody decided to start one.

Who runs them, and where the money comes from

Almost all of them run on short term grant funding, volunteer time or small charges to participants, and very few are commissioned by anybody with a recurring budget.

The money comes from a rotating cast: a local authority small grants pot, a dementia charity, a National Lottery community fund award, a care home’s activity budget, a church, somebody’s fundraising. A typical award covers a year. The leader is paid for the session and rarely for the arranging, phoning and repertoire work around it, and any group of size depends on volunteers for the tea, the chairs and the lifts. Losing the volunteer who does the lifts closes more groups than losing the funding does.

Some sit inside social prescribing arrangements, where a link worker refers people and a scheme pays for places. That is better funded where it exists, with its own constraint: referral schemes tend to buy blocks of a fixed number of weeks, which fits badly with an activity whose whole point is that it does not end. Arts on prescription covers how those referrals work and who pays for arts on prescription covers where the money comes from and how long it lasts.

None of this is a complaint. It is the context for the evidence section below, because a sector funded in twelve month increments does not produce controlled trials, and that explains far more about the state of the literature than any judgement about whether singing groups are worth studying.

The craft of running a good one

Most of what separates a good dementia choir from a mediocre one is decided before anybody sings, and almost none of it is about musical quality.

Key. Sing lower than you would with any other group. Voices not used regularly for decades sit lower than the printed key of anything published for a choir, and a song pitched to sound bright on a recording will exclude half the room. If people are dropping out on the chorus, it is nearly always the key rather than the memory.

Tempo. Slower than the record, not funereal, and absolutely steady. A predictable pulse is what lets somebody rejoin after losing their place, which people do constantly and should be able to do without embarrassment. Rubato is a way of shutting people out.

Repertoire era. Draw mainly from what the room’s own generation grew up with, which for a group of eighty year olds is not what it was five years ago and will not be the same in five more, so check rather than assume. Mix in a little of the very well known material everybody has absorbed regardless of age, because it gives the newest and most anxious people something to hold. Be careful with hymns, which are strongly held and strongly associated with funerals.

Seating and sightlines. A horseshoe or a broad arc, never rows, and everybody able to see the leader’s face without turning, because people follow a mouth far more than a printed word. Carers next to the person they came with, unless the person would rather they were not. And keep the doors visible, because somebody who wants to leave should be able to see how without crossing the room.

Length. Mine run about an hour, of which perhaps forty minutes is singing, and that is a ceiling rather than a target. A group that ends five minutes early with everybody wanting more is better run than one that ends on time with three people asleep.

When somebody cannot follow. Do not fix it publicly and do not slow the room down for it. Move nearer, sing more clearly towards them, and give them something with a smaller entry point, a repeated line or a hummed part. If they are distressed, leave the room with them and come back, or not, and nobody should be made to feel they have interrupted anything.

The tea break. For a lot of the room this is the point, and it is a mistake to think of it as the interval. Twenty minutes of talking to people who understand the situation without being told about it is, for many carers, the only conversation of that kind in a week. I have shortened the singing to protect the tea more than once and I would do it again.

None of this requires a trained musician, though it is easier with one, and none of it requires expensive equipment. It requires somebody watching the room rather than the music. The practical set up is covered on starting an arts programme in a care setting.

What the hour is for the carer

For a substantial part of any dementia choir, the hour is doing something entirely different from what it does for the person with the diagnosis, and pretending otherwise misdescribes the group.

The carer arrives having spent the week managing, explaining, apologising and losing sleep. In the room, three things happen that do not happen elsewhere. Nobody has to have the situation explained to them, including when something awkward happens. The two of them do something together in which neither is the patient and neither is in charge, which for a couple who have lost their previous way of being together is not a small thing. And for an hour the carer is a person singing rather than a person caring, which several have described to me as the only time in the week they are not being somebody’s memory.

That has consequences for how these groups are run and described. The carer’s place is a place in its own right rather than a courtesy, the group should not push them into a helping role during the session, and the tea break is programme rather than logistics. It also means that a group which quietly keeps welcoming somebody after the person they cared for has died is not being sentimental; it is recognising who the hour was partly for all along.

It also means an honest description of what these groups provide includes carer support, which is a different claim with a different literature behind it, taken up on arts for carers and health staff.

What has been measured, and what has not

There is no effect size for community dementia choirs on this page, and that is a decision rather than an omission.

For one to appear, there would have to be a systematic review or an adequately powered controlled trial of community dementia choirs specifically, reporting a design, a named comparator, a primary outcome, a sample size, a follow up interval, a certainty rating and a year. Nothing of that description is locked in this site’s evidence base. So the honest label is Tier 3, of the not adequately studied kind, not the null kind. Nobody has run the study that would settle it, which is a statement about the research literature and not a statement that singing groups do nothing.

The shape of the surrounding literature explains why. The 2024 systematic review of arts on prescription screened 7,805 records and included 25 studies: 10 qualitative, 7 mixed methods, and 8 quantitative studies using uncontrolled before and after designs. Its own words are that “No Randomized Controlled Trials (RCTs) were identified in the search”, and most interventions ran 8 to 10 weeks². It did report a pooled improvement in wellbeing, and that pooled figure rests entirely on studies with nothing to compare against, which is why it stays Tier 3 here and is never called a trial finding.

Community dementia choirs are not the subject of that review and it is not being used as a proxy for one. What it establishes is the shape of the neighbouring literature: overwhelmingly qualitative, almost entirely uncontrolled where it is quantitative, and conducted in short blocks that do not resemble how the activity actually runs. A field that produces short evaluations of eight to ten week programmes was never going to produce good evidence about an open ended weekly group that people attend for five years, and it has not.

Now the part where I am hardest on my own sector, and on myself. I have never measured anything. I have no attendance data I would defend, no baseline of anything, no follow up on anyone who stopped coming, and no idea what happened to most of the people who left. I have written sentences in funding applications about wellbeing and social connection that were true as descriptions of what I saw and were not findings. The absence of research here is not entirely the fault of researchers. The sector has had thirty years and has mostly produced testimony, because testimony is cheaper, it is what funders ask for, and it is what practitioners are good at.

The consequence I care about is not reputational. It is that a family choosing between a choir, a therapist and doing nothing deserves to know which of the three has been studied, and I cannot give them a number for mine.

Why the dementia review does not transfer

The 2025 Cochrane review of music based therapeutic interventions in dementia says nothing about community singing groups, in either direction, and refusing to borrow from it is the single most important discipline on this page.

The review covers 30 studies and 1,720 participants randomised, reporting depressive symptoms at SMD -0.23 (95% CI -0.42 to -0.04) at moderate certainty, which is Tier 1, and agitation and aggression at SMD -0.05 (95% CI -0.27 to 0.17) at moderate certainty, which is Tier 1, null result³. Both belong to the category the review pooled, which is structured music delivered as an intervention. That is not what happens in a church hall on a Wednesday morning with fifty people and a tea urn.

The temptation runs both ways and both are errors. Nobody in my sector should be quoting the depression finding in a leaflet about a choir, and I have seen it done. Equally, nobody should be telling a family that research has shown singing groups do not help with agitation, because the null belongs to a different and more tightly defined thing. A pooled estimate is about the intervention that was pooled, and a distinct practice needs its own study, which is the position personalised playlists are in too. The full account of what the review supports is on music therapy for dementia.

One related caution. Singing is studied properly in at least one other population, and the respiratory work with breathlessness has its own outcomes and its own trials, covered on singing for lung health. Those findings are about breathing in people with lung conditions. They do not travel to dementia either.

Funding fragility, and what happens when a grant ends

The commonest reason a dementia choir stops is not that it failed. It is that a twelve month grant reached month twelve.

That matters more here than for a ten week course, because the whole value of an open ended group is continuity. People with dementia rely on the room being the same room, in the same place, at the same time, with the same faces. A group that stops in March and restarts in September under a different funder in a different hall has not been paused, it has been ended and replaced, and a good number of the original members will not make the transition. I have lost people that way and no funding application captures it.

Three things reduce the risk, none of which solves it. Keep running costs low enough that the group can survive a gap on a small charge and a collection tin. Build the volunteer base wider than one person, so the lifts and the tea do not depend on a single retirement. And keep something written down about how the group runs, so a change of leader is a handover rather than a restart.

Against a background of 57 million people worldwide living with dementia in 2021⁴, the fact that provision of this kind rests on annual grants and volunteer drivers is a policy observation rather than a local misfortune. It is also, circularly, why nobody has the money to study it.

How to find one, or start one

Start with the local dementia support organisation, the memory service, the adult social care listings and the social prescribing link worker at the GP surgery, and then go and watch one before you commit anybody to it.

Watching matters because these groups differ so much. Half an hour at the back tells you whether the leader is watching the room or the music, whether people who lose their place can rejoin without fuss, whether carers are treated as participants, and whether the tea break is twenty minutes or five. Those four things predict whether somebody is still attending in six months better than the repertoire does. Often there is nothing running locally, and starting one needs a hall, a leader who can hold a tune and a room, a kettle and about a dozen people. Arts in care homes and creative activity in dementia care cover the equivalent work inside a home, which is a different job with different constraints.

The last thing is what I would want said to my own family. A choir is not treatment and is no substitute for a diagnosis, a care plan, a medication review or a therapist, and anybody offering it as one is doing you a disservice. It is an hour a week in which two people can do something together that neither of them is failing at. I cannot show you a number for that, and after five years of watching the second row I am not going to pretend the absence of one settles anything.