AbilityBenchAbsolute pitch
Perfect pitch test with no reference tone
Thirty-six single tones arrive one at a time and you name each one, with nothing played beforehand to measure them against — free, no account, and the score is on screen the moment the last tone ends. The octave and the waveform are redrawn every trial and a short atonal run plays in the gaps, because a test that repeats one octave in one timbre is passed by ordinary relative pitch within six trials. Your accuracy is reported against the 8.33% floor with an interval on it, and against how far off you were in semitones.
- 100% free
- No signup
- 36 tones
- Octave and timbre randomized
- Scored against 1 in 12
36 tones, one at a time, each in a randomly chosen octave and one of four waveforms. Name the note. Nothing is played to compare against — not at the start, not between trials, not once — and that absence is the entire test.
The run, in full
- Stimulus
- one tone, 1.2 s long, with nothing played before it to measure it by
- Range
- C3 to B5 — three octaves, with the octave redrawn every trial
- Timbre
- sine, triangle, square, sawtooth, drawn per trial
- Answer
- one of twelve pitch classes; the octave is not asked for and not scored
- Window
- 8 s from the tone starting, then the trial closes unanswered
- Between trials
- 6 atonal tones, to overwrite the pitch you just heard
- Floor
- 8.33% exact, 25% within a semitone
The floor a perfect-pitch result has to clear before it means anything, and over a short run it takes a surprising number of correct answers to clear it convincingly — which is why a result here is reported with an interval rather than as a bare percentage. Scoring that gives partial credit for a near miss raises the floor, and a page doing that recomputes it.
The practice tones have no clock on them and no answer is revealed afterward. That is deliberate: being told the name of a note you have just heard gives you a reference for the rest of the run, which is the one thing this test is built to withhold.
How to take the perfect pitch test
Set the level on noise, name thirty-six tones, read two numbers.
Set your level against pink noise, not a note
The level check plays pink noise rather than a tone, and that is not an aesthetic choice: any note played before the first trial is a reference, and a listener who has just set their volume against A440 has been handed the exact thing this test removes. Start at the low default and raise it until the hiss is clearly present. Headphones are better than speakers here because a small speaker changes the timbre of a square wave enough to make two waveforms sound like one.
Take four practice tones that tell you nothing
The practice tones have no clock and no feedback whatsoever — you will not be told what any of them were. That is the only honest way to offer practice on this particular task, since naming one note out loud gives you an anchor for every note that follows. They exist so that the piano row and the response window are familiar, and nothing else.
Name each tone, then read the two figures that matter
Tap a note on the piano row or press its key: A W S E D F T G Y H U J runs C to B, the layout every tracker uses. The result leads with how many of the thirty-six you named exactly right and the interval around that proportion, then how many landed within one semitone, then your mean error around the twelve-note circle. Anything from 7 correct upward is above what guessing produces in one run out of twenty.
Technical specifications
| Trials | 36 scored tones — each of the twelve pitch classes three times — plus four optional practice tones that are excluded from every figure and carry no feedback |
|---|---|
| Stimulus | one 1.2-second tone per trial, drawn from C3 to B5 so a pitch class never returns at the same height, in one of sine, triangle, square or sawtooth with each waveform level-corrected so brightness is not also loudness |
| Anti-anchor measures | four of them, all stated rather than hidden: the octave is redrawn, the waveform is redrawn, six atonal tones play in every gap between trials, and the response window closes before a hummed reconstruction can finish |
| Response | twelve pitch classes on a piano row, by tap or by key. The octave is neither asked for nor scored — an octave error is the commonest error made by people who do have absolute pitch, so counting it would penalize the ability being measured |
| Window | 8 seconds from the tone starting, which is roughly seven seconds after it stops. A window that closes with no answer is scored as wrong rather than dropped, because eight seconds of silence on a recognition task is an answer |
| Chance floors | 8.33% for an exact name and 25% for landing within one semitone, since three of the twelve answers sit within a semitone of any given note. Guessing at random produces a mean error of exactly 3.00 semitones |
| What counts as evidence | 7 of 36 correct is the point where the exact one-sided binomial probability under guessing drops below 5%. Below that the run cannot separate an ability from a lucky afternoon, and the page says so rather than printing a bare percentage |
| What leaves the page | nothing, unless you press copy, which writes a plain-text summary and a run token to your own clipboard. The token carries the seed and the four figures, so the same thirty-six tones can be rebuilt; it carries no answers |
Frequently asked questions
Why is there no reference tone anywhere, not even before the first trial?
Because a reference is the difference between this test and a much easier one. Given a single named note, most trained musicians can work out any other note by interval — that is relative pitch, and it is common. Absolute pitch is naming a tone with nothing to compare it to, so a page that plays A440 at the start, or lets one note sit in your ear from a level check, has quietly converted itself into a relative pitch test that reports the wrong conclusion. The level check here uses pink noise for exactly that reason, and the four practice tones are never named.
Does the octave count? I knew it was a D but not which D.
The octave is not asked for and not scored, and that follows the research rather than being a concession. Octave errors — naming the right pitch class in the wrong register — are the characteristic error of listeners who unambiguously have absolute pitch, because what the ability delivers is a pitch class rather than a frequency. Scoring the octave would therefore mark down the exact population the test is trying to find. The tones are still drawn across three octaves, but only so the same pitch class never arrives at the same frequency twice.
Why does the waveform keep changing between tones?
Because a fixed timbre gives away pitch through brightness rather than through pitch. Every waveform has its own spectral slope, so within a few trials a listener can start reading how bright a tone sounds and mapping that onto how high it is — a strategy that works within one timbre and collapses the moment the timbre moves. Rotating sine, triangle, square and sawtooth removes it. The four are also level-corrected against each other, since a sawtooth at the amplitude a sine is comfortable at is roughly eight decibels louder and would make brightness a loudness cue instead.
What are the fast tones between trials for?
They overwrite the pitch you just heard. Auditory memory for a specific frequency is good enough to survive several seconds of silence, which means a listener who names one tone correctly can carry it forward and treat the next tone as an interval from it — after which the run stops measuring absolute pitch entirely. Six atonal tones scattered across the full three-octave range, with no key and no repeated pitch, leave nothing to anchor to. They are scheduled on the audio clock in a single call rather than as six timers, because six timers on a busy page arrive unevenly enough to sound like a phrase, and a phrase is a musical object with a key.
Eight seconds is not long. Why is there a limit at all?
Because the slow answers are the ones the test has to exclude. Absolute pitch is a recognition, and people who have it answer in about a second, in the way you name a color rather than the way you work out a sum. The strategies that produce a correct answer without it — humming down to a note you can play, imagining the opening of a song you know the key of, singing a scale in your head — all take several seconds, and a page with no limit collects them and scores them as though they were the same thing. Eight seconds is generous enough that a genuine recognition never gets cut off and short enough that a reconstruction usually does.
Can this page confirm that I have absolute pitch?
No, and the direction of that limit is the opposite of what people expect. A high score fails to rule absolute pitch out, which is genuinely useful, but confirming it needs conditions a browser cannot create: nobody watching that an instrument is not within reach, no tuning app in the next tab, no chance to hum quietly into your own ear before answering. Nothing here can see any of that. A low score, meanwhile, rules out nothing at all — it is equally consistent with small speakers, a noisy room, or a piano row you were still learning.
How many out of thirty-six actually means something?
About 7. Guessing at random over thirty-six twelve-alternative trials produces three correct on average, and the spread around that is wide enough that five or six happens often. 7 correct is where the exact one-sided binomial probability of doing at least that well by chance falls under one in twenty, and the page prints that probability rather than making you take its word for it. It also prints a Wilson interval on your proportion, which is the honest way to say how much of a thirty-six-trial score is sample size.
What absolute pitch actually is, and why nobody agrees how common it is
Absolute pitch is the ability to name a pitch class without a reference, and the operative words are pitch class and without a reference. It is not unusually acute hearing: people who have it resolve small frequency differences about as well as anyone else, and the difference is one of naming rather than of sensitivity. It is not a single ability either — some possessors name notes instantly and lose accuracy for black keys, others are reliable on the timbre of their own instrument and much worse on a sine wave, and a large group sits in the middle with errors clustered one semitone away from the answer rather than scattered. That last pattern is the one worth watching in the table this page prints. Errors bunched at a single semitone describe a pitch template that reads slightly sharp or flat; errors spread evenly across the row describe guessing, and the two produce very similar-looking percentages.
The reason no prevalence figure appears on this page is more interesting than the figure would have been. What is well established is that the ability tracks two things enormously: the age at which musical training began, and whether the speaker’s first language is tonal. A study comparing conservatory students in the United States and China found differences of a size that makes any single population number meaningless — the same test, the same age of onset, wildly different rates. So this page scores against the chance floor instead of against a population, which is a comparison it can actually defend. The relative pitch test is the companion measurement and the one most people should take second: it is this task with the reference put back in, run on the same twelve alternatives against the same floor, so the two scores are directly comparable and the gap between them is the informative quantity. If naming notes is not what you came for but pitch is, the tone deaf test asks whether you can hear a note move at all, which is a different question with a different floor.
Three things go wrong in most browser versions of this test, and all three make it easier in ways that look like nothing. Playing every tone from one octave lets a listener learn the range and then answer by relative height. Using one waveform throughout turns brightness into a usable cue. And leaving the run untimed collects reconstructed answers — hummed, sung under the breath, worked out from a remembered song — and scores them beside recognitions. Take those three away and scores fall a long way, which is the point rather than a complaint. If you want to hear how far pitch judgment can be pulled away from frequency altogether, the Shepard tone illusion separates the two components this test relies on: chroma, the pitch class you are naming, and height, the register it sits in. And if you arrived because somebody told you absolute pitch is a marker of general intelligence, it is not one, and the practice IQ test is the page for that question instead.
A figure this site will not print: absolute pitch prevalence in the general population
The figure usually quoted rests on old, small and self-selected surveys, and prevalence turns out to depend enormously on early musical training and on whether the speaker's first language is tonal — so a single population number conceals the only interesting thing about the distribution.
On the training and language findings: Deutsch, Henthorn, Marvin & Xu (2006), Absolute pitch among American and Chinese conservatory students: prevalence differences, and evidence for a speech-related critical period, Journal of the Acoustical Society of America
Chance level for naming a pitch
Probability of naming a randomly chosen pitch class correctly by guessing: 8.33%.
Twelve pitch classes in equal temperament, so one guess in twelve is right
This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish absolute pitch. Only a qualified professional, working with more than a browser, can make that judgment.
Where the thirty-six answers are counted
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.
The tones are generated by an oscillator in this tab rather than downloaded, so no request goes anywhere while the run is going and the test keeps working with the network off. Your twelve-way answers sit in a JavaScript array, are counted once, and are gone when the tab closes — which is why reloading loses a finished run and why the copy button exists.