Pitch ABX Test — Tuning in Cents

This blind pitch test measures how small a detuning in cents you can hear against a 440 Hz reference, using randomized double-blind ABX trials. All tones are synthesized locally in the browser with the Web Audio API, and nothing is uploaded.

How to use it

1. Choose the detuning rung. The presets are 25, 15, 10, 5 and 2 cents, and the session opens at 15 cents. Changing the rung resets the trial count.
2. Learn the references. Signal A is a 440 Hz tone built from a fundamental plus second and third harmonics. Signal B is the same tone multiplied by 2^(cents/1200), so only its pitch changes. Both are two-second loops with short raised-cosine fades.
3. Listen to the unknown. Listen Unknown X plays a copy drawn at random from A or B on every trial. Nothing in the interface reveals the draw.
4. Make the forced choice. Answer X is Signal A or X is Signal B. There is no “no difference” option, because the two references always differ by the rung you selected.
5. Watch the statistics. Trials, score, accuracy, streak, p-value and the significance badge update after each answer, and the last ten trials are listed in the history row.
6. Walk the ladder downward. Start at 25 cents and move down only when a rung reaches significance on two separate days. Reset between rungs to keep the statistics clean.

The cents arithmetic

One cent is 1/1200 of an octave and 1/100 of an equal-tempered semitone, so two pitches separated by n cents have a frequency ratio of 2^(n/1200). Solving for the interval gives cents = 1200 * log2(f_detuned / f_ref), which is the formula this test uses to place B above A.

Rung	Frequency at A4	Offset from 440 Hz	Beat rate if summed	Musical context
2 cents	440.51 Hz	+0.51 Hz	0.51 Hz	Elite intonation territory
5 cents	441.27 Hz	+1.27 Hz	1.27 Hz	Fine orchestral tuning
10 cents	442.55 Hz	+2.55 Hz	2.55 Hz	Typical studio tuning tolerance
15 cents	443.83 Hz	+3.83 Hz	3.83 Hz	Noticeable on sustained chords
25 cents	446.40 Hz	+6.40 Hz	6.40 Hz	A quarter of a semitone
100 cents	466.16 Hz	+26.16 Hz	26.16 Hz	One semitone, for reference only

The beat column is what you would hear if A and B were played at the same time. The tool plays them one at a time, so the beat rate tells you the size of the difference rather than giving you a cue to listen for. ISO 16:1975 fixes A4 at 440 Hz and asks instruments to hold it within 0.5 Hz, which is about 2 cents at this frequency. The hardest rung here sits exactly at the tolerance the tuning standard itself allows.

Harmonics do part of the work. The 2-cent rung moves the fundamental by 0.51 Hz and the third harmonic by 1.53 Hz, and the ear gets more information from the whole stack than from the fundamental alone. Frequency discrimination is roughly proportional to frequency in the midrange, which is exactly why cents is the right unit for this test and Hz is the wrong one.

What pitch discrimination thresholds look like

Frequency discrimination is usually described as a Weber fraction, Δf/f. Untrained listeners are commonly reported around 1%, which is about 17 cents at 440 Hz. Trained musicians are often placed between 0.2% and 0.5%, roughly 3.5 to 8.7 cents at this frequency, and the best string, brass and keyboard players reach a few cents or better on sustained material. The exact number depends on duration, level, register, and how much the listener knows what to attend to. A two-second harmonic tone with short fades is an easy stimulus by laboratory standards, which is why 2 cents is this ladder’s hardest step rather than an impossible one.

Two perceptual notes matter for how you read a result. Detection gets harder when the two tones are far apart in time, so quick alternation between A, X and B helps. And the frequency region matters as well: pitch cues are most acute in the midrange, which is where the 440 Hz reference sits.

Conditions and scoring

The protocol is the same as every drill in the blind series: fixed references, a hidden X drawn at random each trial, forced choice, and a one-tailed binomial comparison against 50% guessing. At alpha = 0.05 the decision boundary is:

Trials	Minimum correct	p-value	Verdict
5	5	0.0313	significant
10	9	0.0107	significant
10	7	0.1719	not significant
16	12	0.0384	significant
20	15	0.0207	significant

The trial count is open-ended; you decide when to stop, and the badge requires a minimum of 5 trials before significance is even possible. One worked example: at 16 trials, 12 correct gives P = 2517 / 65536 = 0.0384. Seven out of ten is a common result that does not pass, which is the scoring working as intended rather than a verdict on your ears.

Listening conditions decide whether the number means anything:

• Use headphones in a quiet room. Speakers add room color, and a noisy room masks the harmonic information that makes small cents differences audible.
• Keep the level fixed. The two references are identical in level, so a loudness difference points to system processing rather than to the test. Disable spatial audio, equalizers and normalization.
• Do not hum an anchor or check a tuner. Humming a reference tone turns the task into relative pitch, and a tuner display answers the question for you. For this drill, both count as cheating.
• Test with fresh ears. Skip sessions when congested, after loud exposure, or when tired; those states lower measured sensitivity without saying anything useful about your ability.
• Work in short blocks. Ten to fifteen trials per rung, then a break. Pitch judgments drift as attention fades.

Accuracy, limits and troubleshooting

• Sequential comparison is memory-dependent. You are comparing the current tone with the one you heard seconds ago, not hearing both at once. Fast A/X/B/X alternation reduces the memory load and is a trainable skill in itself.
• Pitch is set numerically, not quantized. The oscillator frequency is computed per sample from the cents value, so the rung’s precision is limited by the math, not by the audio device’s sample rate.
• Speakers blur the cue. Room modes can reinforce or cancel the harmonics unequally, which is one more reason headphones are the reference setup for this test.
• A significant result is not a trophy. It means you can hear that difference under those conditions. It does not mean you have absolute pitch, and it does not rank your gear.
• Fatigue produces false negatives. A rung you passed yesterday can fail today on tired ears; always repeat before drawing a conclusion.
• This is educational measurement, not a hearing test. It cannot diagnose anything. If you are concerned about your hearing, see an audiologist.

Related tools

• Blind Frequency ABX Test — the same discrimination task measured in Hz at 1 kHz.
• Tuning Generator — reference pitches including A440 and alternate standards.
• Blind Level ABX Test — check whether a loudness cue is influencing your pitch decisions.