White vs Pink Noise

Both are described as even, and both genuinely are — measured with different rulers. This is also the only pair here that has ever appeared in the same clinical trials, pooled together and never compared.

Two Definitions of Even

White noise has equal power in every hertz. Take any one-hertz-wide sliver of the spectrum, anywhere from the bottom of hearing to the top, and it contains the same amount of energy as any other. That is a completely even distribution, and it is where the word white comes from.

Pink noise has equal power in every octave. The band from 100 to 200 Hz, the band from 1 to 2 kHz and the band from 8 to 16 kHz each carry the same total. That is also a completely even distribution. It is just even according to a different ruler.

The two rulers disagree violently, and the reason is that octaves get wider as you climb. The octave from 10 kHz to 20 kHz spans ten thousand hertz. The octave from 100 Hz to 200 Hz spans one hundred. Under white noise's rule, the top octave therefore holds a hundred times the energy of that low one — and, in fact, roughly half of the signal's total power sits in that single top octave alone. Under pink noise's rule they are equal, which requires the level per hertz to fall as frequency rises, and the rate at which it has to fall works out to 3 decibels per octave.

Human hearing works far closer to the second ruler than the first. We resolve pitch logarithmically, in something like octaves, which is why pink noise strikes most listeners as balanced and white noise strikes them as tilted toward hiss even though white is the flat one. Everything else on this page follows from that mismatch.

The Measured Numbers

Both signals here were characterised over 120 seconds of output at 48 kHz using Welch's method and a least-squares fit between 100 Hz and 10 kHz. White came out at 0.00 dB per octave, ±0.08 dB. Pink came out at −3.07 dB per octave, straight to within ±0.12 dB from 50 Hz to 6 kHz — the most accurate of the five colours generated here.

The divergence between them is therefore 3.07 dB per octave, which compounds to roughly 10 dB at 1 kHz, 16 dB at 4 kHz and 20 dB at 10 kHz when the two spectra are anchored together at 100 Hz. Across the full ten octaves of the audible range the two shapes differ by about 31 dB end to end.

There is a family relationship underneath the arithmetic: pink noise is made from white noise. Every colour on this site starts as white and is shaped by a filter chain, and pink is white tilted down by 3 dB per octave. That is worth knowing because it means nothing has been added — pink noise is white noise with the top progressively taken off, and no other difference exists between them.

Both are normalised to −20 LUFS under ITU-R BS.1770-4 K-weighting, with 0.6 dB of spread across the whole set, so any perceived difference when you switch is a difference in spectrum rather than in level.

The One Practical Difference That Is Not About Taste

Everything else here is preference. This part is physics, and it decides the choice whenever there are voices involved.

Speech carries its loudness in the vowels and its meaning in the consonants, and consonants live mostly between 2 and 4 kHz. Masking works by spectral overlap, so a masker that is strong in that band does not merely make a conversation quieter — it makes it unintelligible, which is the outcome that actually stops your attention locking onto it. White noise is at full strength there. Pink noise, anchored at the same point in the bass, is roughly 15 dB down at 3 kHz.

In practice that means white noise takes a conversation through the wall from words to murmur, and pink noise takes it from words to quieter words. Some people do not care — for them the presence of a voice is the problem and reducing it is enough. For anyone who finds themselves involuntarily following the sentences, only the brighter colour solves it.

The same logic applies to sharp broadband events: a door, a keyboard, a bark, cutlery. White noise covers all of them evenly. Pink noise covers the lower half of each of them, which is usually enough to soften the event and not enough to hide it.

Running the other way, pink noise's advantage in the bass is real but modest. It has about 10 dB more low-frequency emphasis than white at 1 kHz, which helps against a distant road and does very little against genuine rumble. If low-frequency intrusion is the problem, neither of these is the right colour and the comparison you want is a different one.

The Only Pair Here With Trials — Pooled Together, Never Ranked

This is the point that makes this particular comparison worth writing, because it is missing from almost everything written about it: no study has ever compared white noise against pink noise. The literature has treated them as one thing.

Nigg and colleagues (Journal of the American Academy of Child and Adolescent Psychiatry, 2024) is the strongest paper in this area, and its question was, in its own title, whether white noise or pink noise helps task performance in young people with ADHD or elevated attention problems. The two were pooled as a single exposure category. The finding — a small benefit on laboratory attention tasks in the ADHD group, and a small negative effect in comparison groups without ADHD — belongs to the pair jointly. The design cannot say which colour carried it, and the authors do not claim it can.

Riedy, Smith, Rocha and Basner (Sleep Medicine Reviews, 2021) did the same thing on the sleep side, by design: its subject was continuous white noise and similar broadband noise, which is a category, not a colour. Across 38 included articles it rated the quality of evidence that continuous noise improves sleep as very low under GRADE, observed that this contradicts the widespread use of such noise, and raised the possibility that continuous noise may negatively affect sleep and hearing.

The one well-known study that used pink noise specifically is Papalambros and colleagues (Frontiers in Human Neuroscience, 2017), and it is not a comparison either. Thirteen healthy adults aged 60 to 84 slept with a real-time algorithm firing short pink noise pulses at the predicted upstate of their own slow oscillations; overnight word recall improved against sham, while slow wave activity across the whole night was similar between conditions. The stimulus was a click timed to a brain rhythm. White noise was not tested against it, and neither was continuous pink noise.

So when a page tells you the research shows pink noise is better for sleep than white, or the reverse, it is describing something no published study has measured. On this pair the acoustics are the only firm ground, and the acoustics say: brighter masks consonants better, darker is more comfortable for longer. Nothing here is medical advice.

Volume Is the Variable That Actually Matters

The most concrete published measurement anywhere near this topic is not about spectra at all. Hugh and colleagues (Pediatrics, 2014) measured 14 infant sleep machines played at maximum volume: every one exceeded 50 A-weighted decibels at 30 cm, and three exceeded 85 dBA — a level that, sustained beyond eight hours, would exceed adult occupational limits for accumulated noise exposure.

That study is about what devices are capable of when turned all the way up, not about whether a spectrum is white or pink. Its lesson generalises exactly one way: the quantity to be careful with is level multiplied by duration. Choosing between these two colours is a comfort and effectiveness decision. Setting the volume is the decision with consequences.

The practical version is short. Raise the level only until the sound you are covering stops being distinct, then stop — masking is a threshold effect and additional volume past that point buys nothing you can perceive. Put the source across the room rather than beside your head; distance is free attenuation and it takes the edge off white noise more than it takes the body out of pink. And judge the setting lying down in a dark room at the hour that matters, because a room's ambient level falls substantially overnight and a level that felt gentle at ten o'clock is a different thing at three.

Which One, When

Pick white for voices, offices and short bounded sessions. It is the only colour on this site that reliably takes speech from audible to unintelligible, which is why it is the standard choice in open-plan rooms and consulting rooms. It also reproduces almost fully on small speakers, since nothing important is happening in its bass.

Pick pink for nights, for long stretches and as a default. The high-frequency energy that makes white noise effective is what makes it tiring, and fatigue is a function of hours. Pink is the colour most likely to still be running a month later, and the one people describe as sounding like weather rather than like equipment.

Use pink as your reference point when exploring. It sits in the middle of the set, so listening to pink and then stepping one way to brown or the other way to white makes the direction of your own preference obvious within a minute — much faster than trying colours at random.

If you are choosing for work rather than sleep, the evidence such as it is points at this pair. Both were in the 2024 meta-analysis, and both showed the same small effect on lab tasks in ADHD. It gives you no reason to prefer one over the other, but it does mean this is the pair to be experimenting with rather than the darker colours.

The Verdict

Three decibels per octave is a small number that produces a large perceptual difference, because it is applied over and over across the spectrum. White noise is the sharper instrument: better at speech, better at sudden sounds, harder to tolerate for hours. Pink noise is the better companion: comfortable indefinitely, natural-sounding, and adequate at most of what white noise does well without excelling at any of it.

If you have a specific sound to defeat, choose white. If you want something running, choose pink. And distrust any source that claims the research settles it, because the research pooled them and never asked.

Both are generated continuously in the browser here rather than looped, so neither develops a repeat point over a long session, and free sessions run 30 minutes. The free Android app carries both without a session cap, with a sleep timer and offline playback — useful here because the difference between these two really only shows up somewhere past the first hour.

Frequently Asked Questions

Related pages

• Brown Noise for Sleep — The free web player, plus the full guide to brown noise for sleep, focus, and tinnitus.
• White Noise — The flat reference colour, measured at 0.00 dB per octave — the best of the five at making speech unintelligible.
• Pink Noise — Equal energy in every octave, measured at −3.07 dB per octave — the natural-sounding default and the audio industry's calibration signal.
• Brown vs White Noise — The widest gap on the site — 5.98 dB per octave, about 38 dB apart by 8 kHz — and what each colour can and cannot cover as a result.
• Green vs Pink Noise — Green noise here is pink noise plus a 500 Hz bell with both ends cut off — exactly what that chain adds, what it costs, and when the trade is worth it.
• Brown Noise for ADHD — The 2024 meta-analysis with its real numbers — a small benefit on lab attention tasks in ADHD, a negative effect without it, and zero studies of brown noise.