Explanation
Segmentation is the division of a continuous speech signal or sign sequence into units such as sounds, syllables, morphemes, words, utterances, or sentences. These boundaries are not always present as visible cuts in the signal. Spoken words often flow together without pauses, manuscripts may lack word separators, and the same sequence can permit several analyses. Perceivers therefore use prosody, phonotactics, semantic and grammatical expectation, and familiar patterns. Scientific annotation likewise segments for an explicit purpose; phonetic, morphological, and conversation-analytic units need not share boundaries. Segmentation is consequently not neutral preparation but a hypothesis about which parts count for an analysis. The movement leads from an apparently uninterrupted stream to distinguishable units and then back to the question of which relations the imposed cut has revealed or concealed.
Movement of thought
1. A temporal or graphic stream of language first appears continuous and offers several possible sites for relevant boundaries. 2. Prosody, distribution, graphic signs, grammar, and context provide cues about which stretches may belong together as units. 3. Perception or analysis places provisional boundaries and thereby makes sounds, morphemes, words, or utterances comparable. 4. A counter-test shifts boundaries and checks consequences, separating segmentation from pause, blank space, and mechanical division. 5. The divided stream returns as interpreted structure; every cut reveals units while changing their possible relations.
Scientific use case
Phonetics, psycholinguistics, morphology, and corpus annotation compare boundary cues and inter-annotator agreement. Alternatives are retained when different analytic levels require different segmentations.
Example
The sequence “a nice cold hour” may resemble “an ice-cold hour” in rapid speech; lexical and contextual expectations help listeners select boundaries.
Counterexample
An audio file is mechanically divided into exact ten-second intervals without regard to linguistic units. Such technical chunking is not linguistic segmentation.
Distinction
Distinguish it from parsing, tokenisation, word boundary, pause, prosody, morphemic analysis, and mechanical file division; specify unit, purpose, and uncertainty.
Effect on thought and perception
Segmentation makes language data perceptible, countable, and comparable, but premature boundaries may erase ambiguity, conceal continua, or predetermine later analysis.
Resonance
A boundary appears in the flowing signal and units emerge with it; the stream still reminds us that every cut also crosses a connection.
Etymology
Technical core term; individual etymological review pending
Provenance
ATLAS 241 residual-case triage → profile candidate 241-C → source and neighbourhood cross-check → ATLAS 244
Cite this entry
Reich, Michael (Hrsg./ed.): „Segmentation“. Atlas der Sprache, AT-CORE-244-011. ATLAS 260, data model 1.6.0. https://atlasoflanguage.com/concept/AT-CORE-244-011The address remains stable; editorial content may evolve. Editorial reading state: 2026-10-15. Include your access date for precise references.
Historical background
The analysis in Sapir’s second chapter repeatedly changes the unit into which speech is divided. An expression can be examined as a sentence, as a sequence of words or as a construction from smaller meaningful elements. Sapir warns against assuming that the units obtained at these levels are interchangeable. His own examples and field reports lead him to give wordhood a psychological standing, but that conclusion remains his historical interpretation. The chapter is a conceptual witness for linguistic segmentation, not a first attestation of the present technical label or an experimental model of online speech perception. For the Atlas, its value lies in making the analyst’s question explicit. Dividing at grammatical boundaries is not the same operation as detecting acoustic transitions or separating written characters. A division may be useful for one purpose and misleading for another. Later methods of measuring perceptual segmentation and disputes over linguistic units need separate sources; they cannot be retrospectively assigned to this 1921 account.
How conceptual history changes the analysis
Segmentation is level- and purpose-dependent; speech signal, writing, morphology, and syntax need not yield the same boundaries.
References for the historical section
- Edward Sapir, Language: An Introduction to the Study of Speech (1921) · Chapter II, isolating words versus radical and grammatical elements
Original English text read in the Gutenberg transcription. No first-coinage claim, full-book verification or automatic endorsement of the author’s 1921 language classifications and generalisations. Examples require modern descriptive checking before reuse as current language data.