Music Therapy for Depression and Anxiety in Adults: Where the Evidence Sits

Adult mental health is the use most people assume music therapy is best evidenced for, and on this site it is one of the emptiest rows: no effect size for adult depression or anxiety as a primary population is locked here at all. That is a statement about what has been verified for this page, not a verdict on whether the work helps anybody. Both halves of that sentence matter, and the rest of this article is mostly an argument for holding them at the same time.

I run a weekly community singing group, and people join it for low mood. They say so at the door, they say so when somebody asks what brought them, and a good number of them tell me afterwards that it helps. I believe them. I do not treat what they tell me as evidence, and the reasons are not modesty. The people who come are the people who could get out of the house that morning; the ones for whom it was not helping stopped coming and I never learned why; nobody says the group was a waste of time to the person who runs it; and the weeks people describe as their best are often the weeks something else in their life improved. Every one of those is a reason my impression of my own group is close to worthless as a measurement, and I would say the same about anybody else’s impression of theirs.

There was one afternoon that changed how I run it. A woman told me, cheerfully, that this was her therapy now, and when I asked what she meant she said her community mental health team had discharged her in the spring. Within a fortnight I had heard versions of the same sentence from two other people in the room, and a phone call established that a local service had closed a group at the end of its funding and given people a list of community activities, mine included, on the way out. Nobody had done anything wrong and nobody had lied. A door had closed and a leaflet had been handed over. What I did was narrower than I would have liked: I took every word suggesting wellbeing or support off my own publicity, I say out loud at the start of each term what the group is and is not, I keep the NHS and crisis routes pinned up by the door, and I ring the surgery when somebody tells me they have nothing else. That is not a solution. It is the most a singing group can honestly do.

Every claim below is graded per outcome and per population: Tier 1 supported, Tier 2 promising but limited, Tier 3 practice and anecdote. The definitions and the reasoning behind them are on reading arts in health research.

The assumption, and where this page actually stands

Most people arrive at this subject believing adult mental health is where music therapy has its firmest evidence. It is a reasonable belief and it does not survive contact with the sources.

The assumption exists for good reasons. Music therapy is a psychological therapy delivered by practitioners registered alongside other mental health professions, and the intuitive case for it is strongest exactly here: everybody has had the experience of music changing how they feel, so a clinical version of that seems like the obvious application. It is also the version that appears in journalism, in fundraising and on television, and the images that circulate are of adults in distress rather than of somebody working on step cadence in a corridor.

What is actually locked on this site is different. The strongest music therapy finding anywhere here is about walking speed after stroke. The next firmest is about anxiety in the hours around an operation. Neither is about adult mental health, and for depression and anxiety as a primary population there is no effect size on this page, deliberately. For one to appear, this site would need a systematic review or an adequately powered randomised controlled trial reporting a named depression or anxiety outcome, in adults for whom that condition was the reason for taking part, with an effect size, a confidence interval, a certainty rating, a sample size and a year, and readable in the original rather than through a report quoting it. That has not been checked and locked here, so the label is Tier 3, of the kind that means not adequately verified rather than the kind that means the controlled evidence is null.

The preoperative anxiety finding, and what it is not

The 2013 Cochrane review of music interventions for preoperative anxiety pooled 26 trials and 2,051 participants and reports anxiety 5.72 STAI-S units lower (95% CI -7.27 to -4.17)¹. Tier 1, with no certainty rating attached, because the review predates routine certainty assessment and none can be invented for it afterwards.

That is a real finding and a well behaved one. It is also a finding about state anxiety: how anxious somebody feels at a particular moment, in a specific and extremely stressful situation, measured on a self completed questionnaire shortly before a procedure. State anxiety is what the instrument was built to capture, and a hospital corridor on the morning of an operation is the situation it captures best.

A clinical anxiety disorder is a different object. It persists, it is not tied to an identifiable event, it is assessed and diagnosed rather than rated on the spot, and it is treated over months by other means. Moving somebody’s questionnaire score in the hour before surgery says nothing about whether a course of therapy shifts a disorder that has been present for years. Borrowing the figure across that gap keeps the number, the interval and the Tier 1 label while quietly swapping the population and the condition underneath them. You will meet the borrowing, usually in good faith, and it is the precise error this site exists to stop. The finding itself, including why the certainty rating is missing and what the trials actually delivered, is on music before surgery.

The dementia finding, and what it is not

The 2025 Cochrane review of music based therapeutic interventions in dementia reports depressive symptoms at SMD -0.23 (95% CI -0.42 to -0.04) at moderate certainty². Tier 1, and a small effect: roughly a fifth of a standard deviation.

This is the other figure that arrives on pages about adult depression, and the swap is subtler than the first one because the outcome sounds identical. It is not. The review measured depressive symptoms in people with dementia, using instruments designed for that population, in trials run in dementia care settings, mostly with people who were also living with cognitive impairment, often in residential care, and frequently not able to consent for themselves in a straightforward way. Depression as a primary diagnosis in a working age adult with capacity is a different condition in a different life, and the trials that produced this estimate did not include such people.

There is a second reason to be careful with this one. The same review reports agitation and aggression at SMD -0.05 (95% CI -0.27 to 0.17) at moderate certainty, which is Tier 1, a null result, and it found no evidence of any effect persisting four weeks after treatment ends, which is Tier 3 of the nothing adequate has been demonstrated kind rather than the demonstrated null kind. A page that lifts the depressive symptoms row into a claim about adult depression is generally lifting it away from those other two rows as well. The whole table, read together, is on what the dementia evidence shows.

Borrowing either of these findings is not a small stretch made by careless people. It is the standard way the claim gets built, and both borrowings are common enough that a reader searching this topic will meet them within a page or two.

How a mental health singing claim gets built

The clearest worked example in the field is the 2018 three arm trial of singing interventions for postnatal depression, because everything that goes wrong in the retelling can be checked against the trial itself.

Whole sample first, because that is the order the evidence came in. Across the whole sample there was no significant effect, at p=0.16. The significant finding sits in a moderate to severe subgroup at week 6, and it was no longer significant by week 10. The trial had an active comparator, creative play, and singing did not beat it³. The fair summary is faster recovery among the more severely affected women rather than a better endpoint, and it is Tier 2 here on exactly two criteria: the result is a subgroup, and it is an early timepoint.

Now watch how a headline gets made from that. Lead with the subgroup. Omit the timepoint, or mention week 6 without mentioning week 10. Compare against usual care rather than against creative play, which is the comparison that would show the effect was not specific to singing. Describe the whole thing as a randomised controlled trial, which it was, so that the design’s authority attaches to a result the design did not produce. Every sentence in that version can be literally accurate, and the reader ends up believing singing reduces postnatal depression, which the trial did not show.

The three arm design is what makes this checkable, and it is also what makes the trial admirable. An active comparator asks the question that matters: not whether doing something beats doing nothing, but whether this particular something beats another pleasant, structured, sociable activity with other adults in the room. Very few studies in this field are built to answer that, and the ones that are tend to produce less exciting results.

Why this area is unusually hard to study

The evidence here lags behind the intuition for reasons that are methodological rather than mysterious, and naming them is more useful than repeating that more research is needed.

You cannot blind it. Everybody in the room knows whether they are in a singing group. Where the outcome is a self reported mood questionnaire, the participant’s knowledge of which arm they are in feeds directly into the measurement, which is not a problem a larger sample fixes.

The comparator decides the answer. Against a waiting list, almost any structured, sociable, weekly activity looks effective, because it is being compared with nothing happening. Against an active comparator, as in the trial above, the specific ingredient has to earn its place. Trials that choose the first design and report a positive result have often measured the value of turning up somewhere on a Tuesday.

Who reaches a trial is not who is worst affected. Participation requires being well enough to attend, to consent, to complete questionnaires and to keep coming back. The people in the hardest position are systematically less likely to be in the sample, which usually makes the trial population healthier than the population the finding is quoted about.

Therapist allegiance is real. Much of this research is conducted by people who believe in the intervention and often deliver it, which is normal in psychotherapy research and is a known source of inflation.

Follow up is short. Funding cycles produce studies measuring outcomes at the end of the intervention or a few weeks later. Depression is a relapsing condition and the question people care about is what things look like in a year. Almost nothing here can answer that.

None of those obstacles is evidence of absence of effect. They are reasons the file is thin, and telling the two apart is the whole skill this site is trying to pass on. The same pattern shows up next door: dance for depression has an effect size in wide circulation that is quoted with the wrong label and is therefore not repeated here at all, and art therapy for trauma and PTSD meets the comparator problem in a sharper form.

What a session actually involves

Quieter than people expect, and organised around a goal that was written down before it started.

A first session with an adult is usually an assessment rather than music making: what has been going on, what the person wants to be different, what they can tolerate, and how they respond to sound at all. After that, a typical hour might be improvising together on percussion, a piano or a guitar, with the therapist following rather than leading; the therapist playing while the person listens, moves, cries or sits still; working on a piece of music the person has brought because it means something; or writing something, which in adult mental health work is often where the most direct material appears. Nothing has to sound good and no previous musical experience is expected or useful. Musical skill is the therapist’s job. The regulated profession delivering all of this, and the conditions where its evidence is strongest, is set out on music therapy.

Around the hour sits the part that makes it a clinical intervention rather than an activity: written goals, session notes, clinical supervision, liaison with the rest of the team, and a planned ending rather than a course that simply stops. The ending is not administrative. In adult mental health work the shape of the ending is frequently part of the work. Inpatient settings add their own constraints, and those are set out on arts in mental health inpatient units. The line between all of this and what I do on a Tuesday evening is drawn on music therapy against community music.

How people actually get access

Through secondary care, mostly, and less often than the interest in the subject would suggest.

Where music therapy exists in adult mental health it usually sits inside a community mental health team, an inpatient unit, a rehabilitation service or a forensic service, so access follows a referral from a team somebody is already under rather than a request from outside. NHS Talking Therapies services, which handle the bulk of adult depression and anxiety referrals in England, rarely commission it. The workforce arithmetic explains a good deal: the Health and Care Professions Council register listed 6,103 arts therapists across all four protected titles in July 2026, for the whole country and every age group⁴.

Private practice is the other clinical route, and the register is the thing to check before anything else, since using a protected title without registration is a criminal offence rather than a matter of etiquette. The British Association for Music Therapy holds a directory of practitioners⁵, and the practical steps are on how to find a music therapist.

Then there is the community route, which is where most people in this position will actually end up, through a social prescribing link worker or simply by finding something locally. The National Academy for Social Prescribing describes how those referral pathways are meant to work⁶, and the mechanics are on social prescribing explained and arts on prescription. I want to be exact about what that route is. It is a different offer, not a lower tier of the same one. It is unregulated, unassessed, open ended and frequently excellent, and it has no clinical accountability attached to it whatsoever. Being handed it at the point where a service closes is not the same as being referred to it as an addition to one.

Reading the absence correctly

This is the page where the temptation to hear Tier 3 as a refusal is strongest, so the caution is worth stating in full rather than in a clause.

An empty row here means one of two things and this article has said which one each time. It does not mean that music therapy fails for adults with depression or anxiety. It does not mean that a registered practitioner offering it is doing something unsupported, or that a mental health team commissioning it has made an error, or that anybody currently in it should stop. What it means is that no figure has been verified to the standard this site applies, and that anybody quoting one to you should be able to name the review, the population, the outcome and the year.

The reverse error is the one I watch for in myself. It would be easy, having read all this, to conclude that the singing group I run is worthless because nobody has measured it. The people in that room would tell you otherwise, and their view is not evidence but it is not nothing either. What I cannot do is promise anybody it will help, and what nobody can honestly tell them is that it will not. Both sentences have to stay true at once, and the second is the one this field forgets when it is being careful.

If you are unwell, the thing to hold on to is the ordering rather than the tier labels. A creative activity is something to add to whatever a GP or a mental health team has arranged, and not a thing to swap in when a service ends or a waiting list lengthens. When my group became the thing somebody had instead of a service, the honest response was not to run it better. It was to say so, loudly, to the people who had handed out the leaflet.