You listen
Listen to different sounds and silence in turn.
Focus frequency test
Which sound helps you focus?
Try a simple task while listening. Compare your scores with silence.
A verdict takes about 7 runs over several days. If there is no clear difference, the result stays inconclusive.
Headphones recommended Β· No account or payment
Frequency is the difference between the sounds in each ear. This tool does not measure brainwaves.
How a run works
You do a task
Check your reaction time and accuracy with a simple task.
Narrow down, then confirm
Find a candidate in 2 runs, then compare it in 5 more.
The first 2 runs select your candidate. Keep using the same task.
Before you start
Results vary. Silence or a plain tone may work better for you. See the method and evidence
This is not a medical device. Use a low, comfortable volume. Stop if your ears hurt or you feel sleepy.
Getting ready
Getting ready
- Put on headphones. Without them, choose the speaker option below.
- Find a quiet place and silence notifications. Stopping early discards this run.
- There are one silent blocks. They are part of the comparison, not a fault.
- Test: about 7 min Β· 6 blocks Β· 30-second practice first
Settings
The default settings are ready to use.
Task and duration settings
Background sounds turn off during the test. Your focus session settings are kept.
Sound check not done
Listen and set a comfortable volume. You cannot change it during the test.
The tone may sound quieter than music. There is no need to turn it up high.
It is okay if you do not feel a pulsing sensation. If you hear the tone, check left and right below.
Headphone check not done
Listen and identify left or right. Without headphones, choose Speakers above.
Which side did you hear it on?
How the task works
Practice
No sound
60s
0 trials
Block 1 of 7
How absorbed were you during that one?
This is recorded separately from your click scores. Go with your gut β rate the feeling, not how well you think you did.
Not at allFully absorbed
Block 1 of 6 coming up
Take a breath
Drop your shoulders. The next block is a different condition.
10
Practice done
Now the real run. 6 blocks follow in shuffled order. Do not stop partway.
Focus session
Pick a frequency and start the timer
25:00
Idle
Drag around the ring to set the length
Changing the pitch keeps your selected beat frequency.
Background sounds Mix the sounds you like. No frequency setting.
How focused was that session?
Choose what feels right.
ScatteredDeeply absorbed
How did you sleep last night?
Record how last night felt.
RestlessSlept deeply
Sleep mode
Put on a sound before bed
Alpha β delta
The beat gradually drops from 10Hz to 2.5Hz, then fades when the timer ends. A sleep benefit has not been established.
30:00
Idle
Drag around the ring to set the length
Speakers work too. The screen dims after you start.
Keep this tab in front. Locking the screen or switching apps may stop audio. Return and tap the screen to resume.
Background sounds Mix the sounds you like.
Sleep mode: evidence and limits
That sound can affect deep sleep is reasonably well supported. But the evidence is about pink noise, not binaural beats. During deep sleep (slow-wave sleep) the whole brain rises and falls roughly once a second, and pulsing 50-millisecond bursts of pink noise timed to the crests of that wave made the waves bigger and improved memory scores the next day (Ngo et al. 2013, Neuron). The same approach replicated in older adults (Papalambros et al. 2017), and a meta-analysis of 14 studies concluded it raises both slow waves and memory while noting that the studies vary a lot (2023, Frontiers in Sleep).
The heart of that method, though, is reading brain waves live and timing the sound to them. Aimed at the troughs instead of the crests, the effect disappeared. A web page cannot read brain waves, so this screen is not doing that. What it does is play sound regardless of timing, and the evidence for that is much thinner. Still, untimed pink noise has been reported to increase deep sleep in middle-aged adults (2020, Nature and Science of Sleep), so leaving pink noise on in the mixer is not a groundless choice.
Binaural beats and sleep rest on thinner evidence still. A small study found delta-band beats increased deep sleep (N3) (Jirakittayakorn & Wongsawat 2018), as did a pilot on sleep induction (Lee et al. 2019) and a randomised controlled trial following 64 older adults with poor sleep for 14 days (2024). All were positive β but a systematic review covering 2015β2025 concluded that the designs are inconsistent and the samples too small to be confident (2025, Acta Neuropsychiatrica).
Why the frequency descends instead of holding still. Sleep is not one state but a sequence of stages, while most sleep beats on the market sit at a single frequency. That prompted work asking whether stringing several frequencies together in order might track the stage transitions better (2023 study protocol, Frontiers in Neurology), along with a pilot of slowly moving frequencies (2024, SLEEP). Both are protocol or pilot stage, though β neither is a confirmed effect. The 10Hz β 2.5Hz descent here follows that idea; it is not a validated protocol copied over.
So no effect is promised. What this screen does is play a sound that slides from alpha (10Hz) down to delta (2.5Hz), and ask in the morning how it went, then keep the answer. Once those ratings add up, you can judge from your own data whether it helps you.
A/B preference test
Listen to A and B and choose the sound you prefer. Frequencies are hidden. This compares preference, not focus performance.
- Headphones recommended. Binaural mode needs the two ears separated.
- Flip between A and B as often as you like before deciding.
- 6 rounds in all, about 2β3 minutes.
Round 1 of 6
Which one holds your focus better?
Flip between A and B. Replay as often as you want.
Stopping early discards this comparison.
My records
Everything you have logged
Your test, focus, sleep and preference records in one place.
Send anonymous records
When a run finishes, what you see below is sent automatically. It exists to correct this tool's thresholds against real data β the accuracy cutoff that throws a block away (80%), for instance, was borrowed from the literature, and can only be adjusted once we can see what share of blocks it actually discards.
No account, no name, nothing about your device. One random number rides along so that repeat sends from the same browser do not get counted twice. Switching this off also deletes what was already sent.
The only other thing this site counts is page views β no cookies, just the fact that a page was opened (GoatCounter). That is separate from this switch and cannot be turned off.
See exactly what gets sent
β
Frequently asked
Stuck on something?
Start here.
These are the questions people asked after actually running the test. If yours is not here, the evidence page goes deeper into why things are built the way they are.
When something goes wrong
The sound cuts out for a moment. Why?
The cause is usually outside this site. In order of likelihood β β Bluetooth earbuds: the connection wavers, or the buds treat a quiet, steady tone as silence and cut the output to save power. A beat is exactly that kind of sound. β‘ The OS grabs the audio: a notification, a call, another app, or the screen locking. β’ You moved the carrier slider: the sound has to be rebuilt, so it dips briefly and comes back.
There are two things this site can do. Case β‘ is visible to the browser, so it says so on screen, tries to resume, and notes it in the session record. Case β happens past the browser's output and cannot be detected β so the "Sound dropped" button on the session screen marks the moment, and My records shows the pattern: same point every time means power saving; lining up with notifications means the OS. One run with wired headphones settles it.
Can I add it to my home screen like an app?
Yes, and it is recommended. Choose "Add to Home Screen" (mobile) or "Install app" (desktop) in your browser menu and it opens like an app, without an address bar.
On iPhone especially. Safari deletes a site's stored data if you have not visited for about a week, but an app installed to the home screen is exempt. For a tool that needs sessions stacked up over several days, that difference matters.
One catch: on iPhone the installed app may use separate storage from Safari. So the safe order is β β "Save backup file" in the browser you have been using, β‘ add to home screen, β’ open the installed app and use "Load backup file".
Reading the results and the sound
Is a wrong press treated the same as missing one entirely?
For accuracy, yes. Correct presses and correct holds are added up and divided by the total, so a miss and a false press cost the same. They are still recorded separately, though.
The score itself uses only the reaction times of correct presses. A missed trial is left out of the score rather than penalised directly. But if accuracy falls below the cutoff (80% for Go/No-Go and arithmetic, 70% for 2-back), or there are too few correct responses (under 8 for Go/No-Go, 4 for 2-back, 6 problems for arithmetic), the whole block is thrown away β so that something unmeasured is never passed off as measured. That is why a block where you pressed nothing does not count, even if 2-back accuracy comes out high.
The higher Hz sounds louder and more irritating. Are they level-matched?
Every condition is at the same level. The amplitude and the audible pitch (the carrier) are held constant, and only how fast it pulses changes. So it is not a difference in decibels.
It still sounds different because of a phenomenon called roughness. Once the pulsing passes about 15β20 times per second, people stop hearing "a slow swell" and start hearing a buzzing, rough tone. That is why it grates at the same level, and the upper bands here (16 and 18Hz) sit right on that threshold. It is also why the lower bands (6 and 10Hz) sound smoother.
The irritation is left in on purpose β it is part of what is being measured. If 18Hz grates and your score drops, then 18Hz genuinely does not suit you, and the result should say so. It is noted here only so it is not misread as "that one was louder."
Why is the left ear always the lower tone?
No particular reason. No research was found saying which ear should get the higher tone, so one was picked and fixed during implementation. Right now the left is always lower.
Flipping the sides each time would be cleaner in theory: if your two ears differ in sensitivity, that difference currently leans the same way in every block. The comparison between conditions still holds, though, because the silence and pure-tone controls are measured under the same arrangement.
Can the ambient sounds (rain and so on) use my frequency too?
A natural sound like rain has no single frequency to speak of. A sound made of one pitch β the 340Hz tone β can be offset by 5Hz between the ears, but rain is thousands of strands from low to high all at once, so "make the rain 345Hz" is not a statement that means anything.
Playing a beat and ambience together already works, though. Pick a frequency in Focus session and raise rain in the mixer, and both play.
But there is no evidence that the combination adds anything. If anything, studies that mixed binaural beats into noise failed to show an effect. That is why ambience is forced off during measurement. Playing them together in Focus session is for listening pleasure, nothing more.
What this tool cannot do
Does this tell me my brainwave frequency?
No. A website cannot read brain waves. What this measures is not EEG but how fast you perform a task.
The premise that "each person has their own alpha frequency" is well founded, but it can only be measured with EEG. So this site does not claim to find your brainwave frequency. It measures which band raises your score instead. The evidence page has the details.
Is this a medical device? Can I use it as treatment?
No. It cannot be used to diagnose or treat anything.
Keep the volume at a comfortable level β loud listening over long stretches strains your hearing. Do not use it while driving or operating machinery if it makes you drowsy, and stop immediately if your ears hurt or you feel dizzy.
Still not answered?
What is kept here is the background and the limits β things that are hard to leave sitting on a screen. Everything else is answered on the screen where the question comes up. If you want to know why the design is the way it is, the evidence page lays out each decision and its references.
Go measureEvidence
What is known,
and what is not yet.
Which research this site rests on, where the unknowns begin, and how both were folded into the measurement design.
What this test has
It is a 7-minute test running inside a browser, but every safeguard below is in place so the result is not dragged around by chance or expectation. You can see each one at work on the measurement and result screens.
Three things are being blocked β how you happen to feel that day (silence and the pure tone are measured alongside every time), block order (shuffled each run, with the test's permutations built under the same rule), and expectation (you are not told which sound is playing). Each safeguard closes one route by which something other than the sound could reach your score.
Without silence there is no answer to "is sound better than none," and without a pure tone (one tone, no beat) there is no way to separate "the beat did it" from "listening to a sound did it." So both run every time. The largest study in the field, with 920 people, could interpret its negative result on beats precisely because it had silence and tone controls 6.
During a run you are not told what is playing. Someone who has heard "alpha is the good one" pushes harder during the alpha block and gives up early during silence β and that gap is the effect of the label, not of the sound. Every condition is revealed once the run is over.
Block order is randomised every run, which removes order and fatigue effects. Silence is constrained never to come first, and the permutations used by the randomisation test below are generated under that same constraint β if the assignment rule and the test rule disagree, the false-positive rate drifts away from its nominal value.
Through speakers, or one earbud, a binaural beat physically does not form. So this is a gate, not a note: you have to identify left versus right correctly twice in a row before a run will start.
A block whose accuracy falls below the cutoff (80% for Go/No-Go and arithmetic, 70% for 2-back), or that has too few correct responses, is dropped from scoring entirely. Accuracy is not folded into the score because methodological work has shown that combining speed and accuracy into one index usually just inflates variance 7. One score: reaction time. Accuracy is the gatekeeper.
The next sound starts during the break, so the beginning of each block still carries the previous condition. Following N-of-1 trial design guidance, only the later part of each block is scored β the first 20% (up to 10 seconds) is dropped 8.
How you feel changes day to day. The silence and pure-tone baselines have to be measured the same day for the comparison to hold.
A design that sweeps broadly and a design that reaches a conclusion are different instruments. Measuring six conditions a little at a time costs statistical power for no reason other than having many candidates. So the field is narrowed to one candidate first, and only that one runs against silence and the pure tone, two blocks each β tripling the chance of a verdict out of the same 7 minutes.
One or two runs cannot separate a band difference from how you felt that day, so judgement is withheld. A verdict only appears after records have stacked up in the confirmation stage: the condition labels are reshuffled 2,000 times to build a distribution, the observed gap is located within it, and only if a gap this large would arise by chance less than 5% of the time does it read clear difference. This produces fewer false positives than a rule that only looks at the gap between first and second place.
The result screen tests its own assumptions against your data β does the same band keep coming first, do the first and second halves of your records agree on the ranking, is the silence baseline stable, are discarded blocks piling up in one condition. Anything that fails is shown right next to the result. Instead of a trait claim like "your optimal frequency is X," you get a reproducibility figure such as "first in 5 of 8 runs" β because of a known paradox: the more robust a cognitive task, the less reliable it is at measuring individual differences9.
Starting point β alpha frequency differs from person to person
Individual Alpha Frequency (IAF) is a stable trait that varies between people, roughly within 7β13Hz1. The premise that each person has their own rhythm is itself well founded.
There are numbers for how much it varies. In 51 people measured with MEG, the spread between people (SD 2.8Hz) was more than three times the wobble within one person (SD 0.9Hz)2. Individual differences exceed the noise β which is exactly the condition that has to hold for per-person measurement to mean anything.
The same study also shows the limit β IAF differs by about 1Hz between rest and task (parietal, 10.30 β 10.82Hz). IAF is not purely a fixed trait; it moves with state. That is why this site narrows only to bands (theta, alpha, beta) and does not chase 0.5Hz steps. A scale finer than the measurement's own wobble is invented precision.
One citation has been retracted from this page. A 2013 study used to be cited here for the claim that a higher IAF goes with higher general cognitive ability (the g factor β a different g from the effect size g mentioned below). That correlation did not replicate in larger samples: across independent samples of 153 (EEG) and 94 (MEG), fluid reasoning, working memory and g were all unrelated to IAF, with only a weak link to processing speed (about 5% shared variance) 3. So it was removed. What this site leans on is "it differs between people," not "higher IAF is better" β and the former is untouched by that refutation.
But this is a browser β and that has limits
IAF can only be measured with EEG. A website cannot read brain waves. So what this site measures is not your brainwave frequency but how your task performance changes while a sound plays. Those are two different things, and this is as far as a browser reaches.
And the idea that a personalised frequency actually works is itself not yet well established. Meta-analysis puts the cognitive effect of binaural beats at a moderate size overall (effect size gβ0.45 β g measures the gap between two conditions in standard deviations; by convention 0.2 is small, 0.5 moderate, 0.8 large, and negative means the opposite direction), but with very high heterogeneity, so results split widely across people and conditions4. Studies directly comparing personalised against fixed frequencies are thin on the ground.
There are results pointing the other way. In the largest single study in the field (920 people, with silence, pure-tone and classical-music controls), the group listening to binaural beats performed worse6. Even the premise that brain waves entrain to the beat fared poorly in a systematic review: of 14 studies, 5 supported it, 8 went against, and 1 was mixed10. So this site does not assume entrainment. It only measures whether your task performance actually changes β and because results split like this, it measures one person rather than quoting an average.
So it measures instead of claiming
The literature cannot yet say whether personalisation works. So this site runs an A/B test on you alone and shows you the data as it is. Silence and a pure tone are measured alongside as controls, and if silence comes first, it says silence.
Measure it with my own dataHow the literature shaped the design
| What the research says | What was built from it |
|---|---|
| A brief sustained-attention task (PVT-B, 3 min) keeps the sensitivity of the standard 10-minute version5 | Blocks are kept short at 60 seconds, with stimuli packed densely to preserve the trial count (about 35 trials per 60 seconds) |
| "The beat did it" cannot be separated from "hearing a sound did it" without a pure-tone control (our methodological choice) | Both silence and a pure tone are included. If the beat cannot beat the tone, the result says so: the sound itself, not the beat |
| Meta-analysis: masking the beat with white or pink noise is not necessary (similar effect)4 | So pink noise is not used as an active ingredient here. On top of that, a 2024 meta-analysis found cognitive performance got worse for most people (g=β0.21), so it is off by default β turn it on under "pink noise" in the mixer if you want it |
| Exposure before the task (or before and during) beat playing it only during the task4 | The next sound starts during the 10-second break, giving your ears time to settle into it |
| The effect varies widely between individuals (gβ0.45, high heterogeneity)4 | β exactly the premise this product is built on: measure per person |
| In the 920-person study the beats made performance worse β interpretable only because silence and tone controls were present6 | Silence and the pure tone always run alongside, and if silence comes first it is shown as such β treated as a normal result |
| Indices that merge speed and accuracy (IES) mostly just inflate variance7 | One score: reaction time on correct responses. Accuracy is never mixed in β it only acts as a gate |
| In N-of-1 designs, when washout cannot be lengthened, measure only the later part of each period8 | The first 20% of each block (up to 10 seconds) is excluded from scoring |
| The more robust the task, the less reliable it is for individual differences β the reliability paradox9 | Instead of a trait claim like "your optimal frequency," it shows reproducibility ("first in M of N runs") |
| Entrainment itself: 5 of 14 studies supported it, 8 went against, 1 mixed10 | Entrainment is not assumed. Nowhere does this site say it "tunes your brain waves" β it measures changes in performance only |
| A parametric study comparing 340Hz and 400Hz carriers found the lower-carrier combination improved sustained attention11 | The default carrier is 340Hz (changeable in Focus session) |
Superscript numbers refer to the references below.
From the maker
Hunting for focus music, you keep running into "this frequency makes you focus" β and I kept thinking, surely it differs by person, so how can anyone state it flatly? So instead of making a claim, I built something that lets each person measure it with their own data. Here is how it went for me. My first run came back inconclusive β 10Hz (alpha), measured in the small hours, led among the beats but was barely different from the pure tone. Only after stacking up confirmation runs across several days did 10Hz pull clear of silence and settle on clear difference. These days I keep Focus session running at 10Hz. It feels like I get absorbed more easily β and that feeling is recorded separately from the measured score, which is also why this site keeps performance and feeling side by side. This is one person's result. Yours may differ, and silence winning is perfectly normal.
What users said
There is one review here so far. It came from someone I asked to try the tool before public launch, so when and in what setting it was given is stated, along with that person's own result. This site has no review form. Only things said to me in person or in direct contact get posted here, so reviews accumulate slowly.
Turns out silence is my answer. I am someone who needs no distraction at all.
Having several different tasks is a really good idea.
Measured with the Go/No-Go task. Quoted with their approval.
This review is posted as-is for a reason: their result was that the sound did not help. That is a result too, and showing it when it happens is what this site is for. The design criticisms from the same conversation β difficulty drifting between blocks, practice and fatigue effects, and the warning not to generalise from this to studying β were folded into the measurement design and the screens; what changed is recorded in the changelog.
One thing worth saying up front: somebody else's review cannot tell you which band suits you. Large individual differences are where this whole field starts. If you want data across many people, the references below are the better place to look.
References
- Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Research Reviews, 29(2β3), 169β195. doi.org/10.1016/S0165-0173(98)00056-3
- Haegens, S., Cousijn, H., Wallis, G., Harrison, P. J., & Nobre, A. C. (2014). Inter- and intra-individual variability in alpha peak frequency. NeuroImage, 92, 46β55. doi.org/10.1016/j.neuroimage.2014.01.049
- Ociepka, M., KaΕamaΕa, P., & Chuderski, A. (2022). High individual alpha frequency brains run fast, but it does not make them smart. Intelligence, 92, 101644. doi.org/10.1016/j.intell.2022.101644 β the study that refuted the IAFβg correlation from Grandy et al. (2013, NeuroImage), retracted above.
- Garcia-Argibay, M., Santed, M. A., & Reales, J. M. (2019). Efficacy of binaural auditory beats in cognition, anxiety, and pain perception: a meta-analysis. Psychological Research, 83(2), 357β372. doi.org/10.1007/s00426-018-1066-8
- Basner, M., Mollicone, D., & Dinges, D. F. (2011). Validity and sensitivity of a brief psychomotor vigilance test (PVT-B) to total and partial sleep deprivation. Acta Astronautica, 69(11β12), 949β959. doi.org/10.1016/j.actaastro.2011.07.015
- Klichowski, M., Wicher, A., Kruszwicka, A., & Golec, P. (2023). Reverse effect of home-use binaural beats brain stimulation. Scientific Reports, 13, 11079. doi.org/10.1038/s41598-023-38313-4
- Bruyer, R., & Brysbaert, M. (2011). Combining speed and accuracy in cognitive psychology: Is the inverse efficiency score (IES) a better dependent variable than the mean reaction time (RT) and the percentage of errors (PE)? Psychologica Belgica, 51(1), 5β13. doi.org/10.5334/pb-51-1-5
- Kravitz, R. L., & Duan, N. (Eds.), DEcIDE Methods Center N-of-1 Guidance Panel. (2014). Design and Implementation of N-of-1 Trials: A User's Guide. AHRQ Publication No. 13(14)-EHC122-EF. effectivehealthcare.ahrq.gov
- Hedge, C., Powell, G., & Sumner, P. (2018). The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences. Behavior Research Methods, 50, 1166β1186. doi.org/10.3758/s13428-017-0935-1
- Ingendoh, R. M., Posny, E. S., & Heine, A. (2023). Binaural beats to entrain the brain? A systematic review of the effects of binaural beat stimulation on brain oscillatory activity, and the implications for psychological research and intervention. PLOS ONE, 18(5), e0286023. doi.org/10.1371/journal.pone.0286023
- Melnichuk, T., Cooper, K., & Hawk, L. W. (2025). A parametric investigation of binaural beats for brain entrainment and enhancing sustained attention. Scientific Reports, 15, 4308. doi.org/10.1038/s41598-025-88517-z
These references support general conclusions. This site guarantees no particular outcome, and the final word belongs not to the papers but to your own measurements.
Not a medical device; not for diagnosis or treatment. This site uses sound only. Photosensitive epilepsy is a response to visual stimulation, so copying that warning over to audio would be inaccurate β instead, here are the cautions that actually apply. Keep headphone volume at a comfortable level. Loud listening over long periods strains your hearing. Do not use it while driving or operating machinery if it makes you drowsy. Stop immediately if your ears hurt or you feel dizzy.