How to use it
Wear headphones if you can and sit somewhere quiet. Turn the device volume down, press Play reference tone and raise the volume until the tone is clear but comfortable. Pick a base frequency (500 or 1000 Hz suit most people) and press Start practice. In every trial you hear two short tones. Decide whether the second one was higher or lower than the first, then press the matching button or key. If you are unsure, guess.
The measured run starts with a difference of one semitone, which is easy, and makes it smaller each time you answer correctly twice in a row. It ends after 10 reversals or 60 trials, usually within two to three minutes. The result panel shows your threshold and a plot of the whole run.
Controls
- Higher / lower: ↑ or H for higher, ↓ or L for lower, or tap the two large buttons.
- Replay: R plays the same pair again before you answer.
- Esc: stops all sound and pauses the run. A Resume button appears and takes focus; Space or Enter continues.
- Volume: one slider for everything. It has a fixed upper limit and cannot jump up by more than 20 dB at once.
- Base frequency: 250, 500, 1000 or 2000 Hz, locked during a run and remembered next time.
- Stop and show result: ends early. With at least two reversals you get a partial estimate.
How it works
Pitch differences are measured in cents. One octave is 1200 cents and one equal-tempered semitone is 100 cents, so the interval between two frequencies and the size of a difference at a base frequency are:
At 1000 Hz, 10 cents is about 5.8 Hz. The test is a simple direction judgement, so guessing gives 50 % correct. It uses the transformed up-down method described by Levitt in 1971, in its 2-down/1-up form. The difference gets smaller after two correct answers in a row and larger after every wrong one, and the track settles where you are right about 70.7 % of the time. The step is a factor of 2 at first, then the square root of 2 after the second reversal and the fourth root of 2 after the fourth. The estimate is the geometric mean of the last six reversal points; the geometric mean fits because the steps are ratios.
Two details stop you from cheating without meaning to. The first tone is shifted randomly by up to two semitones either way from trial to trial, so remembering one pitch does not help. The second tone is up to 1 dB louder or quieter than the first, so loudness is not a clue. Each tone is a pure sine wave of 350 ms with 20 ms raised-cosine fades, separated by a 250 ms gap. Peaks stay below −6 dBFS, and the signal goes straight to the output with no compressor.
Typical values come from laboratory studies and are only approximate. Well-trained listeners often reach about 0.2–0.5 % (roughly 3–9 cents) at mid frequencies. Many untrained adults reach roughly 0.5–2 % (about 9–35 cents). People with congenital amusia often need changes of a semitone or more. Discrimination also gets worse at high frequencies. Your browser tests use shorter runs and uncalibrated equipment, so treat a single number as a rough guide. A sound card clock error of 100 parts per million shifts pitch by only 0.17 cents, which is small compared with the thresholds measured here.
Things to try
- Run the test at 250, 500, 1000 and 2000 Hz and see where your threshold is smallest.
- Repeat the same base frequency a day later; practice usually lowers the threshold a little.
- Compare headphones with loudspeakers, or a quiet room with a noisy one.
- Try the test when you are tired and again after a break.
- Watch the plot: a clean zigzag around one level means a stable threshold, a drifting track means attention wandered.
- Use Replay on purpose and notice whether hearing the pair twice helps near your limit.
Frequently asked questions
Is this a tone deaf test?
It measures one thing that matters for pitch perception, but it cannot tell you whether you have congenital amusia. That condition is uncommon, with estimates of a few percent of people at most, and it is diagnosed with dedicated batteries that test melody, rhythm and memory as well.
Why do my results change from run to run?
A short adaptive run has random error of a few tens of percent, and attention, practice and noise all play a part. Two or three runs give a better picture than one.
Does the base frequency matter?
Yes. Thresholds in cents are usually smallest in the middle range and grow toward high frequencies, and your equipment may respond differently at 250 Hz or 2000 Hz.
Can pitch discrimination be trained?
Often, yes. Studies report that practice lowers thresholds in many listeners, and musicians as a group tend to have smaller thresholds.
Do I need headphones?
They are not required, but they reduce room noise and make the level more consistent. Loudspeakers in a quiet room also work.