Can an unvoiced sound be loud?

Yes. Lack of voicing does not imply lack of acoustic energy.

Is [s] unvoiced while [z] is voiced?

That is the conventional contrast for these phonetic symbols, although real productions can vary in how continuously voicing is maintained.

Does unvoiced mean whispered?

No. Ordinary speech includes voiceless segments. Whisper is a broader phonation mode and should not be treated as a synonym for each voiceless consonant.

AI Glossary / Natural language and speech / Intermediate

Unvoiced sound

Definition

An unvoiced sound, also called a voiceless sound, is produced without the regular vocal-fold vibration that supplies voicing. In speech, sounds such as a typical sustained [s] can be audible through turbulent airflow even though they lack that voiced source. Unvoiced does not mean silent, meaningless, or outside speech.

What it means

Speech can draw on different sound sources. Regular vocal-fold vibration contributes a periodic component to voiced segments. Air passing through a constriction can instead create the noise associated with a voiceless fricative.
Voicing is one dimension of speech production, separate from where and how the mouth shapes airflow. A sound can be described by its place of articulation, manner of articulation, and voicing characteristics.
Actual speech varies over time. A segment may be partly voiced or become devoiced in context. A phonological category described as voiced or voiceless does not require every recorded realization to exhibit a perfectly uniform acoustic pattern.

How it works

During a voiceless interval, the vocal folds do not provide the regular vibration associated with voicing. Audible energy may come from turbulence or a stop release, depending on the sound’s articulation.
In acoustic analysis, examine cues such as periodicity, noise, and the timing of transitions between segments. Do not equate low energy with unvoiced speech or treat every noisy interval as a speech sound.
Signal-processing systems may assign voiced/unvoiced decisions to short frames. Those operational labels depend on the algorithm, frame length, and recording conditions; they are measurements or predictions rather than direct observations of every articulatory detail.
For phonetic interpretation, combine the signal with context and an appropriate description of the language variety.

Three practical examples

A speech-analysis exercise contrasts a sustained [s] with a sustained [z], listening for the additional voiced component in the latter.
A speech synthesizer models noise-like excitation for appropriate unvoiced portions rather than using a periodic source everywhere.
An audio-processing team checks that its speech detector does not discard quiet voiceless consonants as though they were irrelevant background silence.

Think of it this way

Think of the difference between a steady buzz source and airflow producing a hiss. The analogy helps distinguish voicing from another source of audible energy. It is incomplete because speech combines rapidly changing articulations, voiced sounds can also contain noise, and not every unvoiced sound is a sustained hiss.

Worked example

Imagine a simplified recording of a sustained [s] followed by [a]. An analyst marks the [s] interval from 0 to 120 milliseconds and the vowel interval from 120 to 300 milliseconds.
In this invented illustration, the first interval contains frication without a regular voicing pattern. The second contains approximately periodic cycles spaced 8 milliseconds apart. The corresponding illustrative fundamental frequency is 1/0.008, or 125 Hz.
There is no meaningful 125-Hz vocal-fold fundamental frequency to carry backward across the unvoiced interval simply because an algorithm draws a continuous pitch line. The first interval remains audible speech without that periodic source.
The durations and cycle spacing are teaching values, not measurements from an actual recording. Real boundaries and periodicity can be less clear, especially near transitions or in noisy audio. A failed pitch estimate alone is not sufficient to identify a segment as voiceless.

Illustrative case study

In a fictional project, Willow Speech Lab tests a cleanup filter on short spoken instructions. Reviewers notice that the first consonant of some words becomes harder to hear after processing.
The team inspects both voiced and unvoiced portions and discovers that its simplified rule favors periodic energy too strongly. It revises the evaluation examples to include fricatives and stop releases, then listens to complete words and phrases.
The imagined case demonstrates why speech is not reducible to a pitch track. It does not report a real filter result or imply that all unvoiced energy should always be preserved without considering background noise.

Common misconceptions and mistakes

Unvoiced does not mean silent. Frication can carry substantial audible energy without regular vocal-fold vibration.
It does not mean the speaker has no voice or is unable to speak.
Voiceless and aspirated are not synonyms; aspiration concerns an additional timing and airflow distinction.
A missing pitch estimate can result from noise, algorithm limitations, or other signal properties, so it is not conclusive evidence of voicelessness.
Letter names and spelling do not reliably classify the phonetic realization of a sound in every language or context.

Business use

The concept matters in speech recognition, synthesis, compression, and audio enhancement. Systems that treat speech as exclusively periodic can mishandle consonants that distinguish words.
Evaluate complete linguistic content as well as signal-level measures. Include representative microphones, background conditions, speaking styles, and language varieties. When using a voiced/unvoiced detector, document what its label means and review uncertain transitions. A phonetic label can help explain a processing choice, but the business outcome still depends on intelligibility and task accuracy, not merely cleaner-looking waveforms.

Check your understanding

A noisy but recognizable [s] has no clear periodic pitch track. Must it be removed because it is not voiced?

Answer: No. It can be a meaningful voiceless speech segment. Lack of voicing does not make it silence or unwanted noise.

Related terms

Fundamental frequency
Voicing
Fricative
Aspiration
Devoicing
Speech activity detection

Learn next

Compare voiced and voiceless fricatives before studying stops and aspiration. Learn to distinguish a periodic source from noise-like energy in a spectrogram, and examine transitions rather than assuming each segment is acoustically uniform. Then investigate how a particular voiced/unvoiced detector defines and evaluates its decisions.

Sources

Macquarie University phonetics teaching materials, Distinctive Features, source and voicing distinctions: https://www.mq.edu.au/faculty-of-medicine-health-and-human-sciences/departments-and-schools/department-of-linguistics/our-research/phonetics-and-phonology/speech/phonetics-and-phonology/distinctive-features Macquarie University, Fibrescope demonstration descriptions of vocal-fold behavior: https://www.mq.edu.au/faculty-of-medicine-health-and-human-sciences/departments-and-schools/department-of-linguistics/our-research/phonetics-and-phonology/speech/physiology/movies/fibrescope Independent source review: Macquarie’s source-feature and voice/aspiration definitions and fibrescope descriptions of voiceless stops and fricatives were read; linked movies were not inspected.

Previous: Unsupported-claim rate Next: Up-and-down design