A loud sound masks a weaker sound close to it in frequency

What it is

When two sounds occur at the same time, one loud and one weak, and their frequencies are close together, the weak one may simply not be heard at all — not in the vague sense of being "drowned out," but because its detection threshold is literally raised, sometimes by more than the weak sound's own level, pushing it from audible to inaudible. This is masking, and this entry covers its most basic case: simultaneous presentation with close frequencies.

This is not the same as "hearing a mixture but being unable to tell which sound is which." Masking asks whether the weak sound can be heard at all — a detection question. Attribution — which source a heard sound should be credited to — is a different question entirely.

Why it happens

The cochlea's frequency resolution is not infinitely fine; a given stretch of the basilar membrane responds to a band of frequencies, not a single point. A loud sound's response spreads into neighboring frequency regions, more so toward higher frequencies than lower ones. If a weaker, frequency-close sound is trying to produce a distinguishable response on that same stretch of membrane, its signal gets buried in the "background activity" the loud sound already created there — the weak sound has not physically vanished, but the neural response it evokes is too small relative to the activity the loud sound has already stirred up for the system to pull it out separately, which is functionally equivalent to a raised detection threshold.

The larger the frequency difference between the two, the less their activated membrane regions overlap, and the weaker the masking; the larger the level difference between the loud and weak sound, the wider and deeper the masking extends.

Studying it

The standard paradigm is masked-threshold measurement: a masker (fixed frequency and level) is held constant while a target tone's frequency and level are varied, finding the level at which the target is just barely detectable. Plotting this threshold against target frequency produces a curve that rises sharply near the masker's frequency and falls off on either side.

Typical independent variables: the frequency difference between masker and target, and the masker's level. Typical dependent variables: the target's detection threshold (the elevation relative to its unmasked baseline threshold).

In interface and product sound design research, this method is commonly used to test how much a new cue tone's actual audible threshold gets raised in the presence of existing background sound (system alerts, media playback), to judge whether enough audibility margin remains in a realistic mixed listening scene.

Where it stops holding

• Masking is clearly asymmetric: a loud sound's masking is usually stronger toward higher frequencies than lower ones — the masked region should not be assumed symmetric.
• The effect depends on the level difference between the two sounds reaching a sufficient size; when the two sounds are similar in loudness, noticeable masking does not occur — this conclusion only holds once the level gap is large enough.
• This entry covers energetic masking. Beyond the close-frequency case, a different kind of masking exists that is based on similarity rather than frequency proximity (informational masking), operating through a different mechanism — the frequency-proximity explanation here should not be applied to it.
• How much frequency proximity matters is entirely explained by the spread of basilar-membrane response described above; that explanation is complete here and does not need to be reworked elsewhere.

Applying it

• When designing two sounds meant to play at the same time (an existing cue tone and a newly arriving notification), if the two differ substantially in loudness, pull their dominant frequencies apart into clearly different bands, so the louder one does not raise the weaker one's detection threshold above its actual level.
• Do not rely purely on "turning the volume up" to fix a masked sound — raising the target's level can push it back over the elevated threshold, but at the cost of a noisier overall mix; prefer spectral separation over indefinitely increasing loudness.
• How to check: test the target cue sound in the actual background-sound environment, having participants report the volume at which the target just becomes audible, and compare it against the unmasked threshold measured in quiet — rather than judging "loud enough" from a quiet-room listening test alone.

Related

• Same group: A3.04.2 Masking extends forward and backward in time · A3.04.3 Cue tones need a spectrum offset from ambient noise
• Nearby: A3.15 Auditory stream segregation (this entry answers whether something can be heard at all; that group answers which stream a heard sound belongs to — different questions) · A3.01 Audible frequency range
• Search terms: auditory masking · critical band · masked threshold · simultaneous masking