Tap When You Stop Hearing It Will Flatter You

Morse tells you which bands a phone can carry. This tells you which bands your ears catch. Both have to be true for a high carrier to be useful, and they are different questions. A tone the room can hear is not covert no matter how cleanly it transmits.

Try it in the browser — two beeps, one of which holds a tone. Your browser plays them. Nothing is recorded and nothing reaches my server. Use wired headphones. Through a speaker you are measuring the speaker.

Guessing has to score fifty percent

The obvious design plays tones going up and asks you to tap when they stop. It does not work. People reliably report hearing tones that were never played, and the number you get back is mostly how suggestible you were feeling. Worse, it fails in the flattering direction — the one you least want to be wrong about when the point is whether other people will notice.

So every trial plays two intervals and exactly one contains the tone. You say which. Guessing scores 50% by construction. No amount of wanting to hear something moves that.

Eight trials per band. A band counts as heard only when the score would happen by chance less than 5% of the time, which at eight trials means 7 or 8 correct. 6/8 comes up about one time in seven by luck and is reported as borderline. 5/8 happens more than a third of the time and is not evidence of anything.

Scoring at chance is a measurement. "I didn't hear anything" from a yes/no test is not.

The number used to sit on the card

The frequency used to head each question as "20 kHz". Told the number, people stop attending once it sounds implausibly high. That is an effect on effort rather than on willingness to say yes, so it survives forced choice and pushes the measured ceiling down. Every band is revealed in the results, where it can no longer influence an answer.

Progress is a row of pips, one per band, not a percentage. The ladder stops as soon as two consecutive bands are inaudible, so a bar scaled to the full ladder topped out around 45% and then jumped to the results — indistinguishable from the app giving up on itself.

Hearing is commonly several kilohertz apart between ears. A both-ears test only ever reports the better one. Pick left or right and the tone is panned to that side. A gap of one rung is within the noise of an eight-question test; a persistent gap of two or more across repeat runs is worth mentioning to a doctor, which the page says rather than leaving you to guess.

Two things that throw the whole run out

Fail the 12 kHz check and it refuses to report a ceiling at all. Essentially everyone hears 12 kHz. Failing it means the volume is too low or the output path is not carrying the band — Bluetooth being the usual culprit — and every higher reading would then be measuring that fault.

A tone switched on abruptly begins with a step, a step contains every frequency, and you hear it as a click regardless of the carrier. Test 20 kHz with hard edges and nearly everyone "hears" it. They are hearing the edges. Every tone here ramps over 30 ms, so what is left is the tone or nothing.

A band above half the output sample rate is skipped rather than played. Playing it would not produce silence. It would alias down to something audible, and you would score 8/8 on a band you cannot hear. A false "you hear 21 kHz" is the worst answer this could give.

On a phone, the two answer buttons are 104 pixels tall and get tapped every couple of seconds for about seventy trials. Without telling the browser to treat those taps as taps, Safari reads them as double-tap-to-zoom and the buttons leave the screen mid-test.

Inaudible is a positive finding. "I didn't hear it" is not.

Wired headphones for hearing. The speaker for covertness. Do not mix them up and call both a hearing test.