The ear tells two pitches apart more finely than it tells two loudness levels apart

What it is

Pitch is the perceived highness or lowness of a sound, determined mainly by fundamental frequency. Comparing pitch differences and loudness differences at equivalent proportional change, the ear's discrimination along pitch is markedly finer — making two sounds "sound different" is easier and more reliable by changing pitch than by changing loudness. This doesn't mean loudness differences go unnoticed; at the same relative change, pitch simply supplies more discriminable information.

Why it happens

Whether two stimuli are perceived as different depends on the just noticeable difference (JND) relative to the overall magnitude on that dimension — a ratio called the Weber fraction. The smaller this ratio, the finer the discrimination along that dimension, and the smaller a change needs to be before it's noticed. Pitch's Weber fraction in the mid-frequency range can run far smaller than loudness's, meaning that getting two sounds to be reliably told apart on loudness requires a much larger relative change than doing the same on pitch. This traces back to how the auditory system encodes frequency: the cochlea's basilar membrane responds at different positions depending on frequency (place coding), and different frequencies activate different positions along the membrane, which supports very fine positional discrimination; loudness, by contrast, is encoded mainly through the firing rate of the same population of hair cells, and firing rate has a limited dynamic range, leaving less room for fine discrimination than place coding does.

Studying it

The standard paradigm is JND measurement: present a reference tone, then a comparison tone whose pitch (or loudness) differs slightly from the reference, and use an adaptive procedure (such as a staircase method) to converge on the smallest difference a listener can reliably detect as "different," yielding the JND at that dimension and reference point. The independent variable is the reference tone's frequency or loudness level; the dependent variable is the JND itself and its ratio to the reference magnitude (the Weber fraction). Plotting pitch and loudness JND curves side by side shows directly at which reference points pitch discrimination clearly outperforms loudness discrimination.

This measurement is used to decide, when an interface needs to encode several discrete states along one acoustic dimension, whether pitch or loudness can accommodate more distinguishable steps at the same required perceptual separation.

Where it stops holding

• Pitch's discrimination advantage is most pronounced in the mid-frequency range; the pitch JND degrades noticeably at very low and very high frequencies (detailed in the next leaf), so pitch is not uniformly better than loudness across all frequencies.
• This finding comes from experiments using pure tones or simple timbres; in complex real timbres, pitch and loudness perception interact (a loudness change can produce a slight pitch shift), and actual product sounds are rarely pure tones, so real discrimination performance is more complicated than the lab numbers.
• Hearing impairment, particularly types that degrade cochlear place-coding precision, significantly widens the pitch JND, shrinking pitch's advantage over loudness — this advantage should not be assumed to hold for every user.

Applying it

• When a sound needs to distinguish several discrete states, prefer pitch over loudness as the primary encoding dimension — for the same number of states, pitch encoding can use smaller spacing between adjacent steps and is less likely to be confused under rapid switching or environmental interference.
• Loudness still suits cases with only a few, coarsely distinguished states (such as just "normal" versus "urgent"); it's a poor fit for encoding information that requires fine-grained distinctions.
• How to verify it: build a set of sample sounds for a candidate pitch-encoding scheme and have target users identify which state each corresponds to without seeing a label. If accuracy is noticeably below expectations, the spacing between adjacent pitch steps hasn't met that population's discrimination threshold, and either the spacing needs to widen or the number of steps needs to shrink.

Related

• Same group: A3.11.2 the pitch JND varies with frequency range, sharpest in the mid band · A3.11.3 absolute pitch is rare, but relative pitch comparison is nearly universal · A3.11.4 a melodic contour's rise and fall can be recognized even without an absolute pitch reference · A3.11.5 pitch encoding suits directional meaning, such as a value increasing or a state improving
• Nearby: A3.02 loudness perception and equal-loudness contours · A3.12 timbre and sound quality
• Search terms: pitch discrimination · just noticeable difference · Weber fraction · place coding