Is every stop audibly released?

No. Some realizations have no audible release burst.

Are /p/ and /k/ the same because both are stops?

No. They share a broad manner category but differ in place of articulation.

Is a stop the same as a pause?

No. A stop is a speech sound with an articulatory closure; a pause is a break in the speech flow.

AI Glossary / Natural language and speech / Intermediate

Stop (consonant)

Definition

A stop consonant is a speech sound characterized by a complete closure that blocks the relevant airflow in the vocal tract. In the common oral-stop sense used here, examples include sounds represented by /p/, /t/, and /k/. Their pronunciation involves closure and may include an audible release, but not every occurrence has a clearly audible burst.

What it means

The word stop describes a manner of articulation, not a pause between words. During an oral stop, the lips or tongue form a closure at a particular place, affecting the flow of air and the resulting sound.
Place and voicing are separate properties. For example, /p/ and /b/ are associated with a lip closure, while /t/ and /d/ involve a closure near the ridge behind the upper teeth in common English descriptions.
Actual pronunciation varies with language, accent, position, and surrounding sounds. A phonological category such as /t/ therefore does not guarantee one identical acoustic pattern in every word or speaking style.

How it works

For a typical oral stop, an articulatory closure is formed and held briefly. When the closure is released, the pressure change can create a short burst, sometimes followed by aspiration before a following vowel begins.
The acoustic signal can contain several cues: reduced energy during closure, a release transient, the timing of voicing, and effects on neighboring vowels. Not all cues are present or equally clear in every occurrence.
A speech-processing system analyzes these patterns in context rather than looking only for a burst. Some stops have no audible release, and phonological stops may have other context-dependent realizations.
The term is sometimes used more broadly in phonetic classifications, including nasal stops. This entry focuses on the familiar oral-stop category and keeps nasal airflow and related terminology separate.

Three practical examples

A speech recognizer uses closure and neighboring-sound information when distinguishing words containing different stop consonants.
A pronunciation tool explains that the first sounds in “pat” and “cat” differ in where the closure is made.
A speech-synthesis team checks whether generated consonant timing remains intelligible without exaggerating every release into an unnatural burst.

Think of it this way

Think of briefly closing an air passage and then opening it. That captures the basic interruption of airflow. The analogy is limited because speech includes coordinated tongue, lip, velum, and laryngeal activity; an acoustic recording may not contain a conspicuous pop even when a stop is part of the pronunciation.

Worked example

Consider an invented acoustic annotation of a stop followed by a vowel. The closure begins at 100 milliseconds, release occurs at 170 milliseconds, and regular voicing for the following vowel begins at 210 milliseconds.
The closure duration is 170 − 100 = 70 milliseconds. The interval from release to voicing onset is 210 − 170 = 40 milliseconds, giving a positive voice onset time of 40 milliseconds under this stated measurement convention.
These are different intervals. Calling the entire 110 milliseconds from closure onset to vowel voicing the voice onset time would be incorrect.
The values alone do not establish a universal consonant label, because interpretation depends on language and context. They also do not imply that every stop must have these durations. The exercise illustrates how separate articulatory and acoustic landmarks can be measured without treating one number as a complete description of a speech sound.

Illustrative case study

In a fictional project, Willow Speech evaluates a recognizer on short equipment commands. Some final stop consonants lack a prominent release burst, and an early diagnostic script marks them as missing solely for that reason.
Reviewers inspect the surrounding vowel and closure evidence and revise the annotation criteria. They distinguish an absent cue from an absent consonant and keep uncertain cases available for review.
The imagined workflow shows why a single acoustic feature is insufficient. It does not claim that every reduced sound can be recovered with certainty, or that all accents realize the same phonological stop in the same way.

Common misconceptions and mistakes

A stop consonant is not simply silence. Closure-related low energy occurs within a speech sequence and must be interpreted in context.
An audible burst is not required in every realization.
Letters and sounds are not interchangeable: spelling does not provide a complete phonetic description.
Finally, voicing categories do not mean that every millisecond of one category is voiced and every millisecond of another is unvoiced. Timing and contextual variation matter when examining real recordings.

Business use

Understanding stops helps diagnose recognition and synthesis errors, especially in short words, names, and commands where consonant distinctions can change the intended content. It also supports realistic pronunciation feedback.
Evaluation material should include different positions in words and phrases, natural speaking rates, and the accents the product intends to support. Avoid judging speech solely against an exaggerated demonstration pronunciation.
For annotation, state which landmarks are measured and how uncertain or inaudible releases are handled. A reproducible description of the evidence is more useful than forcing every recording into a simplified pattern that the signal does not clearly support.

Check your understanding

Release occurs at 250 milliseconds and voicing begins at 280 milliseconds. What is the positive voice onset time under the release-to-voicing convention?

Answer: 30 milliseconds. The closure’s starting time is not used in that subtraction.

Related terms

Consonant
Place of articulation
Flap (phonetic)
Phoneme
plosive
voicing
aspiration
voice onset time
flap

Learn next

Compare closure duration with voice onset time, then examine released and unreleased stop realizations. Study how a phoneme can have several phonetic variants without assuming its spelling changes with each pronunciation.

Phoneme

Sources

Sources: University of Groningen, An Introduction to American English Phonetics, stops and manner of articulation: https://opentextbooks.rug.nl/americanenglishphonetics2/chapter/8-4-manner-of-articulation/ . Song, Shattuck-Hufnagel, and Demuth, original study of stop variants: https://www.mq.edu.au/__data/assets/pdf_file/0010/909676/ling-cll-pubs-song-etal-phoneticvariants-2015.pdf . Independently consulted indexed Groningen sections 8.4 and 9.1 on closure, articulation and voicing; direct opening of section 8.4 failed. The original Song and colleagues PDF introduction and section 2.3.1 were directly consulted for unreleased-stop evidence and contextual variation.

Previous: Stoer-Wagner minimum cut Next: Stop list