Using White Noise for Deep Work Sessions

The Origin of the Term

Deep work is professional activity performed in a state of distraction-free concentration. Cal Newport coined the term in 2016 and that definition is the settled part of it: the phrase comes from a trade book rather than a laboratory, so citing it establishes what the term means rather than that long uninterrupted blocks outperform the alternative. No study compares a ninety-minute block against three thirty-minute ones.

What has been measured is narrower and more interesting. The experiments are about interruption, and they do not say what the popular summaries say they do.

What the Interruption Experiments Found

The most-cited laboratory study on this question put forty-eight people through an email-answering task in three conditions: no interruption, interruptions on the same topic as the task, and interruptions on an unrelated topic. The interruptions arrived every two minutes by telephone or instant message.

The result that gets quoted is that interruptions cost you time. That is not what happened. People in the interrupted conditions finished faster than the uninterrupted baseline, with no difference in the quality of what they wrote, and the context of the interruption made no difference at all. The baseline took 22.77 minutes on average; the interrupted conditions took 20.31 and 20.60.

The cost was real but it was somewhere else. After only twenty minutes of interrupted work, the same people reported significantly higher stress, frustration, mental workload, time pressure and effort. The authors' reading is that people compensate for interruptions by speeding up, and pay for the speed in strain.

Two stacked panels. The upper panel plots time to complete the task on a zero to twenty-plus minute axis. The uninterrupted baseline bar runs to 22.77 minutes. The interrupted bar runs to 20.31 minutes and is marked as shorter, crossed by ten tick marks at the two-minute interruption cadence the study used. The lower panel plots three measures reported after the session on the study's one to twenty scale, each as a pair of dots joined by an arrow pointing from the uninterrupted value to the interrupted one. Stress rises from 6.9 to 9.5, frustration from 4.7 to 6.6 and effort from 9.5 to 11.0. A key identifies the first dot of each pair as no interruption and the second as interrupted. Time to complete the task (minutes) 0 5 10 15 20 No interruption baseline 22.77 Interrupted every 2 minutes 20.31 shorter Reported after the session, on the study’s 1 to 20 scale 0 5 10 15 20 Stress 6.9 to 9.5 Frustration 4.7 to 6.6 Effort 9.5 to 11.0 No interruption Interrupted
Mark, Gudith and Klocke's 2008 experiment, from its Tables 1 and 3. The interrupted sessions finished sooner than the uninterrupted baseline, with no quality difference; the cost appeared in what people reported afterwards. The interruption marks are drawn at the two-minute cadence the study used. Scale and values are the paper's own.

Two things follow. First, if you are defending a focus block on the grounds that interruptions make you slower, you are defending it with the wrong evidence. The defensible claim is that interrupted work is more taxing to produce, which is a claim about how the hour feels and what it leaves you able to do next. Second, the effect showed up after twenty minutes. Whatever protecting a block is worth, the measurement happened on a timescale much shorter than the sessions people plan.

A separate observational study of the same group found that work in offices arrives in short segments rather than long ones. It is cited here for that description only; no abstract or full text was available to confirm anything more specific.

The Narrow Job Sound Can Do

Masking does nothing about a tap on the shoulder, a calendar alert or a message that requires an answer. Those are the interruptions the experiments used, and no sound addresses them. What sound can do is narrower: it can cover an intermittent noise that would otherwise pull your attention without carrying any information you need. A door down the corridor is in that category; a colleague saying your name is not.

That is a genuine use, and it is smaller than the usual pitch. It is also not free. A meta-analysis of sixty-five reading experiments found that background noise, speech and music all carry a small but reliable cost, with intelligible speech and lyrics the worst. Silence beats noise. The reason to add sound is that the alternative in your room is not silence, it is someone else's conversation, and a steady sound you chose is the cheaper of the two.

Why the Colour Matters Here

White noise has a flat power spectral density, meaning equal power in every equally wide frequency band. For covering a distraction, that flatness is the useful property: whatever frequency range the distraction occupies, there is energy there. Pink noise rolls off at about three decibels per octave and brown at about six, so both have progressively less energy at the top of the range, which is where consonants and phone alerts live.

The trade is that the same high-frequency energy which makes white noise effective also makes it the harshest of the three to sit under for hours. If white noise is uncomfortable, pink noise is the next option down, and it gives up less speech-range coverage than the slope difference suggests. This site's mixer generates all three, along with green and grey, and can run them together, which is one way to add a little top-end to a mostly low mix.

The Timer on This Site

The mixer page carries a countdown timer with presets of ten, twenty-five and fifty minutes plus a field for any value. When it reaches zero it plays an alarm in repeating groups and vibrates on devices that support it.

Reaching zero also ends the sound. Every active layer ramps down to silence and is left switched off, which means a session has a definite end rather than an alarm over noise that runs until the tab closes. How fast it ends is the fade control's business: with the fade toggle off the layers ramp down over eighty milliseconds, and with it on they ramp over the fade-out duration, thirty seconds by default.

It is still worth being exact about what the timer does not do, because guides describing this kind of tool routinely assume more. There are no interval markers partway through a session, no change in the sound's character during a block, and no record of sessions kept anywhere, so the sound settings and session lengths you used yesterday are not available to compare today.

A separate Pomodoro button automates a different thing: it starts a twenty-five-minute round immediately, fades the sound out at the alarm, counts down a five-minute break on the same clock, and then starts the next round on its own with the colours that were playing before the break. It repeats until pressed again. There is no long break after four rounds and nothing is saved, so the round counter resets whenever the page reloads.

Session Length, Honestly

The figures people repeat for ideal session length — sixty minutes, ninety minutes, the ninety-minute ultradian cycle — are not established by the sources usually cited for them, and this article will not add to them. What the break literature does support is modest and specific. A meta-analysis of twenty-two samples covering about 2,300 people found that short breaks reliably improve vigour and reduce fatigue, with effect sizes around 0.35, but the pooled effect on performance was 0.16 and not statistically significant. Performance benefits appeared only for less cognitively demanding tasks, and longer breaks helped more than shorter ones.

Read that against the thing you are using a focus block for. If the work is demanding — the category where the pooled performance effect was absent — then a five-minute break is at the short end of what has been shown to do anything, and the authors themselves concluded that recovering from highly depleting tasks may need more than ten minutes. The fifty-minute preset with a real break after it has better support than a longer block with a token one.

A Protocol Worth Running

Decide the length before starting, because deciding mid-session is one more thing to spend attention on. Remove the interruption sources you control, since the experiments show those are what cost you, and no sound reaches them. Start the layer you want at the lowest level that covers whatever is intruding, and leave it there — not because consistency conditions a response, which is unmeasured, but because adjusting it is a task and the point is to not have tasks. When the alarm sounds, stop; the sound ends with it, so the block has a boundary you do not have to enforce.

References

Frequently Asked Questions

How long should a deep work session with white noise last?

No source establishes an optimal length, and the commonly repeated sixty and ninety minute figures are not findings. What the break research supports is that breaks help more when they are longer, and that the performance benefit of short breaks was not statistically significant for demanding work. A fifty-minute block followed by a real break has better support than a longer block with a token one.

Do the same sound settings need to be used every session?

There is no evidence that reusing one sound conditions a focus response, so consistency is not worth treating as a rule. The practical reason not to change settings mid-session is simpler: adjusting the sound is itself a task, and the point of the block is to have fewer of them.

Can white noise be combined with music during deep work?

The reading meta-analysis found music carries a cost like noise and speech do, and lyrical music was among the most disruptive conditions tested. If the goal is to cover a distraction rather than to enjoy something, noise alone is the better choice because it carries no information to follow.

What if white noise sounds too harsh for long sessions?

Pink noise is the next option, rolling off at about three decibels per octave, which removes much of the high-frequency content that makes white noise feel sharp. It gives up less speech-range coverage than that slope difference implies. The mixer can also run a mostly brown mix with a thin white layer on top, which keeps some top end without the brightness, or grey, which is shaped to dip where hearing is most sensitive.

Does the timer stop the sound when a session ends?

Yes. The alarm plays and vibrates where supported, and every active layer ramps down to silence and is left switched off. With the fade toggle on it ramps over the fade-out duration, thirty seconds by default; with it off, over eighty milliseconds, which is fast enough to feel immediate and slow enough not to click.

WhiteNoise.top

WhiteNoise.top builds free browser-based noise generators, mixers and breathing timers. Our articles explain the acoustics behind those tools and cite published research where it exists.