The Verification Venue · the ceiling age takes first
The First Note the Years Take
Your hearing nominally runs from 20 Hz to 20 kHz. But the ceiling is the first thing age takes, and it goes on a steep, well-measured schedule. Only about half of healthy 21-to-30-year-olds still hear 20 kHz at all; by 25 most people can no longer hear the 17.4 kHz “mosquito” tone. Below is a live instrument to find your own ceiling by ear, and an honest twist about what that number actually measures.
Every tone here is a pure sine wave your browser generates on the spot: an OscillatorNode whose frequency in hertz is exact to the audio clock. What the readout says is what the speaker is fed. Sweep it upward and, somewhere between about 15 and 20 kHz, it thins to a whistle and then vanishes. That vanishing point is what this page lets you catch.
Audio starts only when you press this. Your browser requires the gesture, and so do we.
Set 1 kHz to a soft, comfortable level. Every test tone is held at or below this level, never louder. If you cannot hear a high tone, that is the answer. Do not turn up.
8.00 kHz · 8000 Hz
The tone ramps softly in and out, never a click. Drag the slider while it plays and watch the spectrum peak below track the number exactly.
Live digital output spectrum · the peak sits on the slider (self-audit)
Arm the tone to see the spectrum. This shows the digital signal leaving the browser, before your speakers touch it: a single pure-tone spike that lands exactly on the frequency shown above.
Sample rate & Nyquist ceiling
–
the hard cap on any frequency your device can carry
Your captured cutoff
not yet found
run the sweep, then confirm it by ear
Press and hold. The pitch climbs slowly. Let go the moment you can no longer hear it. Sweeping up from an audible tone is safer and less suggestible than jumping to a target.
Confirm it: which interval had the sound?
Two short intervals will play in a random order. Only one carries the tone. Pick the one you heard. Beat chance twice and the reading stands.
– your confirmed cutoff
–
A cutoff is a soft gradient, not a wall. A good result does not certify healthy hearing (early loss hides in mid frequencies this test never touches); a poor result on cheap earbuds means little. This is not audiometry.
What your browser is actually working with. Change one thing here, your headphones for your laptop speakers, wired for Bluetooth, and re-run the sweep: your “cutoff” will move by kilohertz while your ears stay exactly the same.
Where your ceiling sits
In a sound booth with calibrated transducers, Wang and colleagues (2021) measured 162 healthy adults aged 21 to 70. Below 16 kHz, every under-30 responded. At 20 kHz only 52.2% of them did, and above 40 essentially no one produced a measurable threshold up top at all. Their median thresholds (in dB hearing level, higher meaning more loss) climb steeply with each decade. Your confirmed cutoff drops onto the same axis as a vertical line.
Median extended-high-frequency threshold by age band (Wang 2021)
Each curve is one age band; a higher line means a worse (louder-needed) threshold. Curves stop where fewer than about 1 in 5 people still responded. The full numbers, including where the response failed, are in the table further down, which is the text version of this chart.
The shape of the decline: ISO 7029 models it as a parabola in age
ISO 7029 gives the median age-related shift as ΔH = α(Y − 18)²: zero at 18, then accelerating, which is why the ceiling marches down the frequency axis faster and faster with age. This shows the shape only (normalized), because the standard’s α values are frequency-specific and were revised in 2024. Crucially, the standard does not reach the band this instrument tests:
The popular “you heard 15 kHz, your hearing age is 40” tables extrapolate ISO 7029 into a band it never measured. That is folklore shaped like a standard. Use the standard for the shape of the decline, not to stamp a number on your birth certificate.
Why the top fades first
This is not random wear. The cochlea is tonotopic: it is a coiled ramp that sorts frequency by place. The stiff, narrow base resonates to the highest frequencies; the floppy apex to the lowest. Age-related loss (presbycusis) is a sensorineural loss that kills the outer hair cells of the basal turn first, and those basal cells supply the most gain, roughly 50 to 70 dB of cochlear amplification at the base versus only about 20 dB at the apex. When they die, the high frequencies do not fade gently. They collapse. The current test tone is mapped onto its place below.
The unrolled cochlea · base (high) to apex (low)
The basal end (left) carries the highest frequencies and the most gain, and is the first to die with age. Your test tone sits near the far basal edge, which is exactly why it is the first note the years take.
What you are really measuring
Here is the twist the gimmick “ear age” tools hide. This instrument measures your loudspeaker, your DAC and your room at least as much as it measures your cochlea. Three honest layers:
1 · Playback roll-off and the Nyquist ceiling
Consumer earbuds, laptop speakers and Bluetooth codecs attenuate hard above roughly 15 to 16 kHz, and the browser’s sample rate hard-caps everything. At the common 44.1 kHz rate the Nyquist ceiling is 22.05 kHz; at 48 kHz it is 24 kHz. A tone near 20 kHz may be reproduced weakly, aliased, or over a lossy Bluetooth link essentially not at all. So “I cannot hear 18 kHz” is ambiguous between a healthy-but-older cochlea and a device that never delivered a clean 18 kHz. That is why the flag lights up near the ceiling, and why switching devices moves your number.
2 · The ISO 7029 standard stops short
The presbycusis ear-age standard, ISO 7029:2017 (amended 2024), tabulates median thresholds only from 125 Hz to 8 kHz for ages 18 to 80, with an extended set merely to 9 to 12.5 kHz for ages 22 to 80, and above 8 kHz the over-70 values are published “for information only.” It says nothing about 15 to 20 kHz. The honest use of it is to explain the shape of the decline (quadratic in age, steepest up top), not to convert your cutoff into a hearing age.
3 · A cutoff is a gradient, and fresh results flatter you
The under-25 versus over-25 mosquito split is a soft population gradient with wide individual scatter, not a hard wall. A good result here does not certify healthy hearing, because early loss can sit in mid frequencies this page never tests. A poor result at 19 kHz over cheap earbuds means almost nothing. This is a physics demo of tonotopy and presbycusis, not a diagnosis. Real extended-high-frequency testing uses calibrated transducers in a sound booth, which is where the Wang 2021 numbers come from.
The ladder, in numbers
The chart above is a picture of this table. “NA” means the response rate was equal to or below 20%, that is, most of the band could no longer hear that frequency at any presented level. Read down a column and you watch a decade of aging; read across the top rows and you watch the ceiling fall.
| frequency | 21–30 | 31–40 | 41–50 | 51–60 | 61–70 |
|---|
Responsiveness under 30: 100% at 16 kHz, falling to 52.2% (12 of 23) at 20 kHz. The 17.4 kHz “mosquito” tone (Stapleton, 2005) sits in the gap that opens up first.
The check: the signal math and every sourced number, recomputed
Nothing here is a stored screenshot. The button below synthesizes pure sines in JavaScript, runs a discrete Fourier transform on them, and re-asserts the physics and the citation live. It mirrors the offline verifier exactly.
The offline gate recomputes all of it, two independent ways where it can: node research/highest-pitch-you-can-hear/verify-highest-pitch-you-can-hear.mjs.
Free choices and uncertainty, named. The Wang 2021 ladder is one healthy cohort, not a universal constant; exact extended-high-frequency thresholds vary by study, transducer and calibration standard. The 20 kHz upper limit is a young-adult nominal most adults never reach. The ISO 7029 curve is shown as shape only (its α coefficients are frequency-specific, paywalled, and were revised in the 2024 amendment), never as a dB reading in the 15-to-20 kHz band. The AnalyserNode spectrum shows the digital signal, so it proves displayed frequency equals played frequency, but it cannot see your speaker’s roll-off, which is the whole point of the device panel. Your captured cutoff is bounded by your hardware and your room, not just your ears.
What is exactly true here, and what is a model
Exactly true (the sourced facts). Nominal human hearing spans about 20 Hz to 20 kHz. The top of the band declines first and earliest with age, because presbycusis kills the basal-turn outer hair cells first and the cochlea is tonotopic (base high, apex low). In Wang 2021, 100% of under-30s responded at 16 kHz and 52.2% at 20 kHz, with median thresholds climbing steeply per decade. The 17.4 kHz mosquito tone is real (Stapleton, 2005). ISO 7029 models the median shift as a quadratic in age and does not cover above 12.5 kHz. The sample-rate Nyquist ceiling (22.05 kHz at 44.1 kHz, 24 kHz at 48 kHz) is exact.
A demo, not a measurement (your number). Your captured cutoff depends on your transducer, DAC, codec, room, background noise and volume as much as your cochlea. It is not calibrated in dB HL, it is not audiometry, and it does not yield a hearing age. The 50-to-70 dB (basal) versus about 20 dB (apical) cochlear-amplifier gain figures are representative values from the pathophysiology literature, illustrating why the top collapses hardest, not a spec of your own cochlea.
Named simplifications. The spectrum is the digital output, before hardware. The age placement is a soft gradient with wide scatter, shown against one cohort. The ISO curve is normalized shape, not dB. The mosquito line is a population tendency, not a threshold of birth date. None of these change the reversal at the heart of the page: the top really does go first, and this browser number really is measuring your gear too.