Ceiling
1 · Set the volume
Play the reference and set your device volume so it is clearly audible but comfortable. Then leave the volume alone — changing it part way through makes the result meaningless.
Why it matters this much
A speaker measures your hearing and that speaker's roll-off together, and small speakers fall off a cliff above 16 kHz — you get a low ceiling that is mostly the hardware's fault. Bluetooth is worse: many links resample and discard this band before it ever reaches you.
Testing through a speaker is still right if the question is “would somebody standing near this device hear it” — that is the path they would hear it through. Just do not call it a hearing test.
Eight forced-choice trials per band. A band counts as heard only when the score beats chance at p < 0.05.