The human ear has a well-defined upper and lower frequency limit

What it is

The frequencies a human ear can perceive are not unlimited — they are compressed into a band running roughly from 20 hertz to 20 kilohertz (20Hz–20kHz), commonly called the audible frequency range. Sound below this lower bound is infrasound; above the upper bound is ultrasound; neither counts as an "audible sound" to a typical adult ear no matter how intense it is. The range is also not equally sensitive throughout: the same sound pressure is easiest to detect somewhere around two to five kilohertz, and the closer a frequency sits to either edge, the more sound pressure it takes to be heard at all — "audible" itself comes in degrees, not as one flat band.

Why it happens

For a sound to be perceived, it first has to pass through the mechanical transmission of the outer and middle ear, after which hair cells in the inner ear's cochlea convert that mechanical vibration into a neural signal — and this chain itself sets the limits on which frequencies can be effectively processed. The cochlea's basilar membrane responds to frequency in a spatially organized way (tonotopic organization): frequencies that are too low fail to produce enough vibration amplitude along the membrane, while frequencies that are too high exceed the physical limits of the hair cells' mechanical response. At both ends, the signal decays below what can be effectively encoded before it ever reaches the nervous system. The mid-frequency band is most sensitive partly because the outer ear canal's own resonance amplifies sound pressure in that range, and partly because the cochlea's mechanical response is most efficient there — together, these two factors make the mid band the lowest-threshold stretch of the entire audible range.

Studying it

The standard method for measuring an individual's hearing range and thresholds is pure-tone audiometry: tones at a series of frequencies are presented, starting clearly audible and stepped down in loudness until the listener reports just failing to hear them, and the lowest audible sound pressure at each frequency is recorded, producing an audiogram. This curve varies from person to person and also shifts with the background noise of the test environment and the actual frequency response of the headphones or speakers used, so the specific threshold numbers it produces cannot be lifted out of their test conditions and reused elsewhere. In interface research the more common practice is not to remeasure a full audiogram from scratch, but to cite the established norms from audiology directly, reserving a dedicated measurement for cases where a specific population's actual hearing (older users, for instance) needs to be confirmed.

Where it stops holding

20Hz–20kHz is a rough range given for young adults with normal hearing, not a precise number that applies to everyone: the upper bound drops noticeably with age (a general pattern in how hearing changes over the lifespan), the lower bound stays comparatively stable, and individual variation can make a person's actual range narrower or wider than this commonly cited figure. It's also worth separating "can be heard at all" from "can be heard clearly enough to use as information" — a sound sitting just above threshold at a given frequency only means it isn't being fully masked; whether it sounds loud, clear, or comfortable depends on the specific relationship between loudness perception and frequency.

Applying it

• Don't design an interface's alert or notification sounds with their frequency content hugging the upper or lower edge of the audible range, and especially avoid putting a critical pitch change in a region that is noticeably high or low, since those are exactly the regions where most people's hearing is weakest.
• Pay extra attention when using small speakers (phones, wearables) — their actual physical frequency response is often narrower than the ear's audible range, so confirm the hardware can actually reproduce the designed sound's frequency band before asking whether users can hear it.
• To verify: play the designed alert sound on the target device and inspect it with a spectrum analysis tool to see where its main energy is concentrated, confirming that band is both within the range most people can hear and within the range the device can actually reproduce — both conditions are required, neither is sufficient alone.

Related

• Same group: A3.01.2 High frequencies are lost preferentially with age · A3.01.3 Alert sounds should not place critical information solely in the high frequencies
• Nearby: A3.02 Loudness perception and equal-loudness contours · A3.19 Auditory fatigue and safe volume
• Search terms: audible frequency range · pure-tone audiometry · absolute threshold of hearing