Research note · Published September 2026

Does music help you sleep?

The best pooled evidence says people who listen to music at bedtime report sleeping better, by a large margin. The instruments strapped to those same people show almost nothing. Both results are real and the distance between them is the interesting part.

This is one of the few questions in my field where somebody has already done the careful version of the work. Cochrane reviews are the closest thing medicine has to a referee, and there's one on exactly this. It was first published in 2015 and updated in 2022, and it pools 13 randomised trials covering 1,007 adults between 18 and 83, run in Taiwan, Iran, China, Singapore, Hungary, Denmark, Italy and Austria.

What moved

The headline outcome is sleep quality as people report it, measured on the Pittsburgh Sleep Quality Index. That's a questionnaire scored from 0 to 21 where a lower number is better sleep, and anything above 5 is the usual line for poor sleep.

Across 10 of the studies and 708 people, the music groups scored 2.79 points lower than the groups who got no treatment or ordinary care, with a confidence interval running from 3.86 to 1.72. The reviewers describe that as an improvement of roughly one standard deviation, and they rate the certainty of it as moderate. In a literature where most effects are small and heavily qualified, that's a big result and a well-graded one.

What didn't move

Almost everything else.

Insomnia severity, measured in two studies with 63 people between them, showed no clear difference at all, and the reviewers rated that evidence very low certainty. Perceived sleep interruption showed no effect. Depression among people with sleep problems showed no effect across three studies and 173 participants.

Three outcomes did shift a little, all of them self-reported and all rated low certainty: how long people took to fall asleep, how long they slept in total, and how efficiently they slept. Each of those rested on three studies and 197 people, which is a thin base for a claim.

Then there's the finding that does the real work in this review. Where the same outcomes were measured objectively rather than by questionnaire, the benefit mostly disappeared. People wearing the equipment didn't fall asleep meaningfully faster or sleep meaningfully longer. The questionnaire moved by a standard deviation. The machines moved by very little.

Reading the gap

There are two ways to take that, and the split is where most of the writing on this subject goes wrong by picking one and not saying so.

The deflationary reading is that these trials can't blind anybody. You know perfectly well whether music is playing. If you were handed a bedtime routine by a researcher and asked afterward how you slept, the expectation alone can produce most of a 2.79. The reviewers name this themselves: participants knew their allocation, outcome assessors weren't always blind, samples were small, and the measurement methods varied between trials. That's why the certainty ratings sit where they do.

The other reading is that subjective sleep quality is a separate thing from the objective numbers rather than a flawed proxy for them, and for insomnia it's arguably the thing that matters. Nobody comes in complaining that their sleep-onset latency is 27 minutes. They come in because the nights feel bad. An intervention that changes how the night feels, without changing the polysomnography, has still changed the complaint.

I hold both. The effect is probably smaller than 2.79 once you account for what people knew, and what's left over still reaches the part of the problem people actually present with.

What the trials asked people to do

This part is more specific than most summaries let on, and it's the part a reader can use.

Participants listened to prerecorded music, on their own, mostly at bedtime. Sessions ran 25 to 60 minutes, averaging about 36. The trials ran anywhere from three days to three months. The music was classical, jazz, new age, lullabies and traditional music, depending on the study.

So the tested thing is a defined block of listening at the end of the day rather than background sound all night, long enough to run past the point where you'd normally lie there. Whether the music matters more than the block of time does, no trial in this set can tell you, because none of them ran a silent 36 minutes as the comparison.

Where this sits

No study in the review reported an adverse effect, which changes the arithmetic for anyone deciding whether to try it. The downside is 36 minutes.

It isn't a treatment and I'm not a clinician. The American College of Physicians doesn't recommend for or against music for insomnia, and continues to put cognitive behavioural therapy for insomnia first. If your sleep has been bad long enough that you're searching this question, that recommendation is the one to act on, and a doctor is the person to take it to.

What the evidence supports is narrower and still worth having. It's thirty-six minutes of music at bedtime, across 13 trials and 1,007 people, with nobody reporting harm, and a questionnaire that came back better in the morning.

Sources

  1. Jespersen, K. V., Pando-Naude, V., Koenig, J., Jennum, P., & Vuust, P. (2022). "Listening to music for insomnia in adults." Cochrane Database of Systematic Reviews, 8, CD010459. DOI: 10.1002/14651858.CD010459.pub3 View source
  2. Cochrane plain language summary, "Music for insomnia in adults." View source
  3. "Insomnia Therapy: Listening to Music." American Family Physician, Cochrane for Clinicians, 2023. View source