Does White Noise Help You Sleep?

Last checked 3 November 2026 · 13 min read

Sometimes — and almost never for the reason the box gives. White noise is not a sedative and it does not make you drowsy. What it does is flatten the difference between the quiet of the room and the sound that wakes you. That is a narrow job, and it explains the whole pattern of the evidence: a large effect in a hospital or on a main road, no measurable effect in an ordinary bedroom, and a small but real effect on how long it takes to fall asleep. It also predicts the two ways it fails — if you can hear the machine, it has stopped being a floor and become an event; and if it switches off at three in the morning, the step it creates is the same size as the one it was hiding.

What wakes you is the step, not the noise

Start with the experiment that is small, old and more useful than most of what has been written since. Five people slept in a laboratory with electrodes on, for three nights each. One night was quiet. One night played a recording of a working hospital ward. The third played the same recording with white noise mixed into it, so that the background was louder but the jumps between background and peak were smaller.

The numbers are the argument. In the quiet night people were woken or half-woken 13.3 times an hour. Under the ward recording it rose to 48.4 — nearly four times as often. Add the white noise and it fell back to 15.7, close to the quiet night, even though the room was now louder than it had been on either of the other two nights.

Why did the louder room produce fewer awakenings? Because the researchers measured the wrong thing first and then found the right one. They had assumed that the loudest peaks did the damage. What actually predicted an awakening was the change from the room's background level to the peak — the size of the jump. And that jump was the same on the ward night and the white-noise night, at 17.7 and 17.5 decibels. The white noise did not soften the door slam. It raised the floor underneath it, so the door slam had less distance to fall.

This reframes the whole question. White noise is not a blanket over the world; it is a floor under it. It cannot stop a sound arriving, and it does not teach you to ignore sounds. It reduces the size of the arrival. Everything that follows — who it helps, who it does nothing for, how loud it should be — comes out of that one sentence.

Three predictions you can check at home

If the mechanism is the step, then the people it helps are not "light sleepers" in general. They are people whose awakenings are triggered by arrivals: a car door, a floorboard, a neighbour's television, a partner turning over, a dog two gardens away. Their nights are quiet enough that each of those is a cliff edge.

The people it does nothing for are the ones whose awakenings are not sound-driven at all. If you wake at three because your body temperature has peaked, because your bladder has filled, because your back hurts, or because your cortisol is doing what it does before dawn, then a fan is a fan. The room is not the problem, so improving the room changes nothing. This is the most common disappointment with white noise, and it is not a failure of the device.

The third group is the one nobody warns you about. If you are the kind of sleeper who monitors sound — who notices the fridge, then cannot stop noticing it — adding a continuous noise gives you one more thing to notice. For those people the machine is not a floor. It is a new event, running all night.

The fan that changed nothing

The best test of the ordinary case is unglamorous and unusually clean. Forty-eight healthy young adults slept at home for two nights, in random order: one with a standard air-conditioner hum at 43 decibels, one without. Light and room temperature were measured at the same time, and sleep was recorded by actigraphy.

Nothing moved. Total sleep time was 467 minutes with the sound and 479 without it. Sleep onset was 10.8 minutes with the sound and 15.1 without — a difference in the right direction that the statistics could not distinguish from chance. Sleep efficiency was 81 per cent against 78.8 per cent. Three outcomes, three null results, in the population most likely to be buying these machines.

That result only looks like a contradiction. In a quiet suburban bedroom there may be almost no steps to blunt: the house is already near its floor, the loud events are rare, and a fan mostly adds a sound rather than removing a jump. Compare it with the study of ten adults in New York City, all of whom had complained that environmental noise was wrecking their sleep. They used a fan-based machine for a week, between two baseline weeks. Their time awake after falling asleep improved on the wrist monitor, and their sleep latency improved on the diary. Same intervention, opposite result — because the room was different.

So the first question is not "does white noise work". It is "what is the size of the jump in my bedroom, and how often does it happen". A flat on a bus route and a house at the end of a cul-de-sac are not the same experiment.

The hospital test that fixed the room and not the patient

The most recent trial is also the most sobering, because it did everything right and still split in two. Twenty-nine older patients in a hospital were randomised to a control or an intervention level of white noise, played continuously from ten at night to six in the morning, and then crossed over to the other condition the following night. Eighteen completed both nights. Sound meters in each room counted the jumps of 17.5 decibels or more — the size that has been shown to raise the odds of waking.

The intervention worked on the room. Those jumps fell from a median of 24 a night to 11. It did not work on the people. Sleep fragmentation on the actigraph was no different between the two nights, and neither were the patients' own ratings of their sleep.

That is the most useful result in this whole literature, and it is almost never quoted, because it is not a yes or a no. It says the mechanism is real and the outcome is not automatic. You can measurably improve the sound environment of a room and still not change a night, because a night is made of more than sound. Anyone selling a machine on the strength of "it blocks noise" is telling you the first half of that sentence and stopping.

How loud, and how quiet

The one trial that produced a clear benefit in healthy sleepers used a level that would surprise most buyers. The quiet condition was a normal room at about 40 decibels. The active condition was broadband sound at about 46 decibels — roughly the volume of a quiet conversation, or a fridge heard from the next room. That was enough to cut the time to fall asleep from 19 minutes to 13. It was not loud. It was just above the room.

That gives a calibration rule better than any decibel chart: the machine should be only just louder than the quietest thing you can hear when it is off. Set it until the room tone disappears, then take it down one notch. If you can hear the machine as a sound — if you can tell it is a fan, or a hiss, or a track — you have raised the floor past the point where it does its job and turned it into the thing your brain is now tracking.

There is a safety floor at the other end, and it matters most for the smallest sleepers. Fourteen infant sleep machines were measured at maximum volume, at 30, 100 and 200 centimetres, with a correction for the size of a six-month-old's ear canal. Every one of them exceeded 50 decibels at 30 centimetres — the level recommended as the limit in hospital nurseries. Three exceeded 85 decibels, which, run for more than eight hours, is past the occupational limit for an adult's daily noise exposure. The study was published in 2014 and the devices are still sold; the volume dial is the part of the product nobody reads.

The machine becomes a sound of its own

Here is the part that follows from the mechanism and almost never gets written down. If what wakes you is a step, then the machine's own edges count. Switching on is a step. Switching off is a bigger one.

People describe this without knowing they are describing it. One sleeps with a pedestal fan set to run for twelve hours and says that if it shuts off, it wakes her. Another has her white noise set to stop at waking time, and reports that she usually wakes a minute or two before it stops — because the volume ramps down first. Both are describing a trained response to the edge of a sound, which is the same response the white noise was hired to blunt.

The practical consequence is simple and slightly annoying: a timer defeats the purpose. A machine that runs all night keeps the floor flat. A machine that stops at 3am creates the largest step of the night, in the half of the night when you are already sleeping lightest and least able to absorb it. If you need it off by morning, ramp it down over twenty minutes rather than cutting it, and expect the first few nights of the ramp to be worse.

Why it keeps some people awake

The failures are consistent, and they are mostly about the sound itself rather than the sleep.

It is the wrong colour. White noise puts equal energy at every frequency, which is why it reads as hiss. Several people describe it as too intense or ear-piercing, and switch to pink or brown noise, which carry more energy at low frequencies and sound more like rain or a distant motorway. This is a preference, not a hierarchy: there is no trial showing one colour beats another for sleep. But if white noise makes you grit your teeth, the colour is the first thing to change, before the volume.

It is a vigilance cue. A sound that never resolves keeps some brains in checking mode. People with tinnitus describe this sharply: the noise that masks the ringing can also become the thing the brain listens to for changes. One described white noise as making her anxious in a way that echoed an old trauma, and said silence did the same — the problem was not the sound but the monitoring.

You needed silence to stay asleep. The split is not noise versus no noise; it is noise to fall asleep versus silence to stay asleep. One person put it exactly that way, and it is a common pattern: the sound helps you cross the threshold and then keeps you in the shallower stages. If your first hour improves and your second half gets worse, the machine is doing its job and costing you more than it pays.

It cannot fix a night that is not about sound. The heaviest users are often the least helped. One person described sleeping inside a wind tunnel of a box fan, a tower fan with a sheet over it and an air conditioner, and still having terrible insomnia. The room was as flat as a room can be. The problem was elsewhere.

A three-night test

Before buying anything, find out whether the step is your problem. It takes three nights and a phone.

If the note by the bed says "sound" three nights running and the fan nights are better in the back half, you have a sound problem and a cheap solution. If it says heat, bladder, pain or thought, spend the money elsewhere — and if it says "unknown" every night, the answer is probably not in the room at all.

Where this fits

Sound is one of four things a bed provides: sound, weight, heat and movement. They arrive together and they do the same job — they tell an old part of the brain that the room is occupied. Remove one and the night changes shape, which is the subject of why you sleep worse when your partner is away: the empty bed takes three of the four at once, and the pillow pile people invent is an attempt to buy back the weight. The bought version of that is in whether a weighted blanket works, which restores pressure and nothing else.

If the machine helps you fall asleep and the night still breaks at three, the problem is not the room: waking at three covers the second half of the night, and waking before the alarm covers the version where the night simply ends. If you are reaching for sound because the silence is unbearable rather than because it is loud, that is a different mechanism, and the page on not trying is closer.