Connected Speech and Coarticulation
Chapter Objectives
By the end of this chapter, you will be able to:
- Distinguish coarticulation from categorical phonological alternation.
- Identify assimilation, reduction, deletion, epenthesis, resyllabification, and linking in controlled data.
- Explain gradient versus categorical change at an introductory level.
- Compare citation forms with connected realizations of the same words.
- Identify style and rate effects on connected speech.
- Produce a controlled connected-speech transcription using only previously taught notation.
An Observable Problem
Every transcription task so far in this book has worked with words spoken carefully, one at a time, or with short phrases treated as fixed wholes. Real speech rarely works that way. Say the phrase "did you" at a natural conversational pace, and compare it with the same two words pronounced slowly and separately, as if reading a word list aloud. For many speakers of many languages, the natural version sounds noticeably different from the sum of its two citation forms: a consonant may sound different at the boundary between the words, a vowel may shrink toward a shorter and less distinct quality, or the two words may seem to blend into something that is difficult to segment cleanly into "did" plus "you."
None of this is disorder, laziness, or a broken version of careful speech. It is systematic, and every fluent speaker of every language does some version of it. The challenge for a phonetician is not to judge these changes but to describe them precisely: what changed, why it changed given the surrounding sounds, whether the change is a small automatic adjustment or a larger categorical substitution, and how a transcriber should represent it responsibly.
This chapter asks three questions:
- What is the difference between a small, automatic, context-driven adjustment to a sound and a larger, more categorical change that a language treats as an alternation between distinct forms?
- What recurring named processes account for most of the differences between a citation form and its connected-speech realization, and how can a beginner identify each one in verified data?
- How do speech style and speech rate affect how much connected-speech change actually occurs, and what should a transcriber do differently for careful, formal speech compared with fast, casual speech?
Coarticulation Versus Categorical Alternation
Purpose: establish the central distinction of the chapter before naming any individual process, so that every later example can be sorted correctly.
Coarticulation is the overlap of articulatory gestures for neighboring sounds, so that the vocal tract begins moving toward a following sound, or continues moving from a preceding one, while a given segment is still being produced. Coarticulation is largely mechanical and gradient: it reflects the physical reality that the tongue, lips, and jaw cannot move instantly between target positions, and the amount of overlap can vary continuously with rate, style, and position rather than jumping between a small number of fixed outcomes.
A categorical phonological alternation, by contrast, is a change that a language treats as a choice between a limited number of discrete forms, typically describable as one sound class replacing another in a definable context. A categorical alternation is often reported consistently by fluent speakers as one sound or another, not as a continuous range in between, and it may be reflected in how the language's own speakers or its established phonological description treat the pattern.
A useful starting test, though not the only evidence a full phonological analysis would use, is to ask whether the observed variation forms a continuous range that tracks rate and context by degree, which points toward coarticulation, or whether speakers and analyses instead describe the outcome as a small number of distinct categories with comparatively sharp boundaries between them, which points toward a categorical alternation. Many connected-speech processes discussed in this chapter can show both a gradient version, which is coarticulation proper, and a more categorical version in fast or casual speech, and part of the transcriber's task is deciding which is supported by the evidence at hand.
This distinction connects directly to the phone-phoneme-allophone framework introduced in Chapter 5. A coarticulatory adjustment is ordinarily allophonic: it does not change which phoneme a listener perceives, only some phonetic detail of its realization. A categorical alternation may still be allophonic, if it is a fully predictable, conditioned substitution, or it may cross into a different kind of process entirely, such as deletion or epenthesis, which is why this chapter treats coarticulation as the starting concept rather than the whole story.
Error: Treating every audible difference between a citation form and a connected-speech form as evidence of "sloppy" or "incorrect" pronunciation.
Correction: Connected-speech processes are systematic and rule-governed, not random degradation. A careful phonetic description states what changed and under what conditions, without characterizing the change as an error.
Citation Forms and Connected Realizations
Purpose: define the two reference points that every comparison in this chapter depends on.
A citation form is the pronunciation of a word produced in isolation, typically slowly and carefully, as if reading a word list or answering "how do you say this word." A connected realization is the pronunciation of the same word as it actually occurs within a longer, naturally paced utterance, where it is influenced by neighboring words, phrase position, prominence, and rate.
Dictionaries and early textbook chapters generally supply citation forms because they provide a stable reference point for teaching an inventory of contrasts, consistent with the broad-transcription policy from Chapter 24 previewed here. Citation forms are not incorrect, and connected realizations are not incorrect; they answer different questions. A responsible transcription task, following the purpose-dependent principle introduced earlier in this book, always states which kind of form is being represented.
Notice that the connected form above is bracketed with square brackets, marking it as a phonetic realization, while the underlying words retain their citation-form phonemic identity. Comparing the two forms side by side, rather than presenting only one, is the method this chapter uses throughout.
Naming the Processes
Purpose: introduce a working vocabulary for the most common connected-speech processes, each defined, illustrated, and distinguished from its neighbors.
Assimilation occurs when a sound becomes more similar to a neighboring sound in one or more features, such as place, manner, or voicing. Assimilation may affect a sound that comes before the trigger, called anticipatory or regressive assimilation, or a sound that follows it, called perseverative or progressive assimilation.
Reduction describes a shift of a vowel, or occasionally a consonant, toward a shorter, less peripheral, or less distinct realization, frequently affecting unstressed function words such as articles, prepositions, and auxiliary verbs. Reduction is closely tied to the stress and prominence patterns introduced in Chapter 20: unstressed material is where reduction is most common, though it is not automatic in every language or every unstressed position.
Deletion, sometimes called elision, is the complete absence of a segment that is present in the citation form, most often affecting an unstressed vowel or a consonant in a complex cluster.
Epenthesis is the insertion of a segment not present in the citation form, frequently occurring to break up a sequence that the language's phonotactics, introduced in Chapter 19, otherwise disfavor.
Resyllabification occurs when a segment is reassigned across a word or morpheme boundary to a different syllable than it occupies in the citation form, typically so that a consonant at the end of one word becomes the onset of a syllable that begins the next word.
Linking is the absence of an expected break between two adjacent words or syllables, so that the boundary between them is not marked by a pause, glottal onset, or other discontinuity. The linking mark introduced in Chapter 22 for intonational purposes is reused here to mark this segmental-level connection.
Real connected speech frequently shows more than one process at once. A single phrase may show assimilation at one boundary and vowel reduction within the same word. Naming a single dominant process in a complex example, when more than one is genuinely present, is a common source of incomplete analysis; state every process you can support with evidence rather than settling for the first one you notice.
Notating Connected Speech
Purpose: consolidate how to represent connected-speech processes using only notation already taught, connecting to the delimiter and diacritic policy established throughout the book.
This chapter introduces no large new symbol set. Instead, it teaches learners to apply previously taught segmental symbols, the syllable boundary . from Chapter 19, and the linking mark ‿ from Chapter 22 to represent the processes above within square-bracketed phonetic transcription.
| Symbol | Name | Use in this chapter |
|---|---|---|
| [ ] | Phonetic brackets | Encloses a connected realization, as distinct from a citation-form phonemic representation in / /. |
| . | Syllable boundary | Marks a resyllabification when the syllable affiliation of a segment changes in connected speech. |
| ‿ | Linking mark | Marks the absence of an expected break between adjacent words or syllables. |
| (∅) | Deletion notation | A conventional way to mark a deleted segment in a working transcription; not an official IPA symbol, and should be labeled as a project convention when used. |
Represent assimilation and epenthesis by transcribing the actually occurring segment, drawn from the symbol inventory already taught, rather than inventing new notation. Represent reduction using vowel symbols already introduced for central, reduced vowel qualities. As with every transcription task in this book, state the reference variety, the style, and the approximate rate before presenting a connected-speech transcription as evidence.
Error: Marking every instance of assimilation, reduction, or deletion in every word of a transcript regardless of whether the surrounding context actually supports it, simply because the process is possible in the language.
Correction: A process being possible in a language does not mean it occurs in every instance. Base each transcription decision on the specific evidence for that token, consistent with the broad-versus-narrow judgment developed in Chapter 24.
Style and Rate Effects
Purpose: show that connected-speech processes are not fixed properties of a phrase but vary systematically with speech style and rate.
The same phrase, spoken by the same speaker, can show substantially different amounts of assimilation, reduction, deletion, or linking depending on speech style, ranging from careful, formal, or clearly enunciated speech to casual, fast, or intimate conversational speech. Formal registers, public reading, and speech directed at an unfamiliar listener tend to preserve more citation-form detail; casual, fast, and familiar registers tend to show more extensive connected-speech change.
Rate interacts with style but is analytically separable from it: a speaker can produce careful speech quickly under time pressure, or casual speech slowly when relaxed. Because both factors affect the amount of connected-speech change, a transcription intended as evidence should document both the style and the approximate rate of the source material, consistent with the metadata requirements introduced for recording-based transcription in Chapter 27.
It is tempting to describe connected-speech processes as switches that are simply on in fast speech and off in careful speech. The evidence instead points to a continuum: many processes show intermediate, partial realizations at intermediate rates and styles, particularly for coarticulatory effects, which are gradient by nature. A responsible transcription acknowledges this continuum rather than forcing every token into one of two extreme categories.
Across Languages
Nasal place assimilation, in which a nasal consonant adopts the place of articulation of a following consonant, is documented across a wide range of unrelated languages and is not a pattern unique to English. The specific consonants affected, the direction of the assimilation, and whether the process is optional or effectively obligatory in ordinary speech vary by language and by the specific analysis proposed for that language, so a general statement about "nasal assimilation" should not be treated as predicting the exact behavior of an untested language without further evidence.
Standard French is well documented as having a pattern often called liaison, in which a word-final consonant that is normally silent in isolation is pronounced when the following word begins with a vowel, linking the two words together in a way that partly resembles the resyllabification and linking processes described above. Liaison is governed by a mix of grammatical, lexical, and stylistic factors specific to French rather than being a simple universal linking rule, and its details are usually treated as requiring dedicated instruction in a French-specific course rather than full derivation from general phonetic principles alone.
Languages that restrict which consonant sequences may occur, particularly in onset or coda position as introduced in Chapter 19, frequently show epenthesis when a loanword, compound, or rapid-speech sequence would otherwise violate that restriction. The specific vowel or consonant inserted, and the exact contexts that trigger insertion, are language-specific facts that must be established from evidence for each language rather than assumed from a general tendency.
Analysis Exercises
These exercises apply the coarticulation-versus-alternation distinction and the six named processes to controlled data. Justify each decision with reference to the citation form, and avoid marking a process that the data do not actually support.
Exercise A — Naming the Process
For each pair of citation and connected forms, name the single best-supported process illustrated and briefly justify your choice.
- Citation "handbag" /ˈhændbæɡ/ → Connected [ˈhæmbæɡ] (General American)
- Citation "and" /ænd/ → Connected [ən], as in "bread and butter" (General American, casual)
- Citation "warmth" /wɔːrmθ/ → Connected [wɔːrmpθ] (General American, some speakers)
Exercise B — Gradient or Categorical?
For each scenario, state whether the described variation is best treated as gradient coarticulation or a more categorical alternation, and explain what additional evidence would strengthen your answer.
- A vowel's tongue position measured across many tokens shows a continuous range of values that shifts predictably with the backness of the following consonant.
- Speakers and a reference grammar consistently describe a nasal consonant as taking one of exactly three place values depending on the following consonant, with no reported intermediate values.
Exercise C — Style and Rate
Predict how the connected transcription of "did you eat" would likely differ between a careful, formal reading and a fast, casual conversational rendering, and identify at least one process likely to appear only, or mainly, in the casual version.
Common Mistakes
Error: Describing connected speech as a corrupted or lazy version of careful, citation-form speech.
Correction: Connected-speech processes are systematic, rule-governed, and present in every fluent speaker's ordinary speech. Careful and connected forms answer different descriptive questions rather than representing correct and incorrect versions of the same target.
Error: Generalizing a connected-speech pattern documented for one variety, style, or speaker to an entire language without qualification.
Correction: State the reference variety, style, and approximate rate for every connected-speech claim, consistent with the reference-variety policy applied throughout this book, and avoid presenting one speaker's casual register as representative of the whole language.
Error: Treating every difference between a citation form and a recording as deletion, without considering assimilation, reduction, epenthesis, resyllabification, or linking as alternative explanations.
Correction: Work through each named process before settling on an analysis, and prefer the explanation best supported by the specific articulatory and acoustic evidence available for that token.
Chapter Review
Part A — Coarticulation and Alternation
1. Explain the difference between coarticulation and a categorical phonological alternation, and describe one form of evidence that helps distinguish them.
2. Explain how this chapter's distinction connects to the phone-phoneme-allophone framework from Chapter 5.
Part B — Naming the Processes
1. Define assimilation, reduction, deletion, epenthesis, resyllabification, and linking, each in one or two sentences.
2. Explain why a single connected-speech example may require naming more than one process.
Part C — Style, Rate, and Variation
1. Explain how speech style and speech rate each independently affect the amount of connected-speech change observed.
2. Explain why a connected-speech transcription should always state the reference variety, style, and approximate rate of its source.
Cumulative Review
1. From Chapter 19: Explain how resyllabification in this chapter builds on the syllable-constituent concepts introduced in Chapter 19.
2. From Chapter 22: Explain how the linking mark introduced for intonational purposes in Chapter 22 is reused for a different, segmental purpose in this chapter.
3. Preview for Chapter 24: This chapter distinguished citation forms, represented phonemically, from connected realizations, represented phonetically. Predict how Chapter 24's broad-versus-narrow distinction will refine this comparison further.
Self-Check
Before moving to Chapter 24, confirm that you can do the following without looking at your notes:
- Distinguish coarticulation from a categorical phonological alternation using at least one form of supporting evidence.
- Define and identify assimilation, reduction, deletion, epenthesis, resyllabification, and linking in controlled data.
- Compare a citation form with a connected realization of the same phrase.
- Explain how style and rate each affect the amount of connected-speech change.
- Transcribe a short connected-speech example using only previously taught notation, correctly bracketed.
- Avoid describing connected speech as a corrupted or incorrect form of careful speech.
If any of these tasks are difficult, review the relevant section before reading Chapter 24. Chapter 24 turns transcription-level choice itself into an explicit analytical decision, building directly on the citation-versus-connected comparisons developed here.
Answer Key (click to expand)
Analysis Exercises — Exercise A
1. Assimilation; the alveolar nasal takes on the bilabial place of the following stop, matching the regressive place-assimilation pattern illustrated earlier in the chapter. 2. Reduction combined with deletion; the vowel reduces toward a central quality and the final stop is deleted, a common pattern for this function word in casual connected speech. 3. Epenthesis; a voiceless stop is inserted between the nasal and the following fricative, matching the pattern illustrated for "something" earlier in the chapter.
Analysis Exercises — Exercise B
1. Gradient coarticulation; a continuous range of measured values that shifts by degree with context is the signature of an automatic, mechanical overlap rather than a choice among a small set of discrete categories. Further evidence could include measuring whether the range is truly continuous across many speakers and rates. 2. A categorical alternation; consistent description of exactly three discrete place values, with no reported intermediate values, points toward a phonological pattern the language treats as a choice among distinct categories rather than a purely gradient effect. Instrumental measurement across many tokens would strengthen the claim by checking for any overlooked intermediate values.
Analysis Exercises — Exercise C
A careful, formal reading of "did you eat" would likely preserve more of each word's citation-form detail, with clearer boundaries between words. A fast, casual rendering is likely to show place or manner assimilation and possible coalescence at the "d" plus "y" boundary, along with vowel reduction on the unstressed "you," illustrating that the amount of connected-speech change is not fixed but varies with style and rate as described in this chapter.
Chapter Review — Part A
1. Coarticulation is the gradient, mechanical overlap of articulatory gestures for neighboring sounds, tracking rate and context continuously; a categorical alternation is a change a language treats as a choice among a small number of discrete forms with comparatively sharp boundaries. Evidence such as continuous instrumental measurement versus a small number of consistently reported discrete outcomes helps distinguish them. 2. A coarticulatory adjustment is ordinarily allophonic, changing phonetic detail without changing which phoneme a listener perceives, connecting directly to the phone-phoneme-allophone distinction from Chapter 5; a categorical alternation may also be allophonic if fully predictable, or may involve a different kind of process such as deletion or epenthesis.
Chapter Review — Part B
1. Assimilation: a sound becomes more similar to a neighboring sound in one or more features. Reduction: a vowel or consonant shifts toward a shorter or less distinct realization, often in unstressed position. Deletion: a segment present in the citation form is absent in the connected realization. Epenthesis: a segment not present in the citation form is inserted, often to repair a disfavored sequence. Resyllabification: a segment is reassigned to a different syllable across a word boundary than in the citation form. Linking: an expected break between adjacent words or syllables is absent. 2. A single connected-speech example can show more than one simultaneous change, such as assimilation at a word boundary together with vowel reduction within the same word, so an analysis that names only one process may miss genuine, evidence-supported changes elsewhere in the same example.
Chapter Review — Part C
1. Style affects how much a speaker adjusts toward careful, citation-like forms or toward more extensive connected-speech change based on formality and audience; rate independently affects the physical time available for articulatory targets to be reached, so a speaker can be careful but fast, or casual but slow, and both factors combine to determine the observed amount of change. 2. Because the same phrase can be realized very differently depending on style and rate, a transcription without this metadata cannot be evaluated or compared reliably, consistent with the purpose-dependent transcription principle applied throughout this book.
Cumulative Review
1. Resyllabification depends on the onset, nucleus, and coda concepts introduced in Chapter 19; it describes a segment moving from one syllable's constituent structure into a different syllable's structure across a word boundary, which requires the constituent vocabulary already established. 2. In Chapter 22, the linking mark indicated the absence of an expected intonational break at the phrase level; in this chapter, the same mark is reused to indicate the absence of an expected segmental break between adjacent words or syllables, extending a previously learned symbol to a related but distinct kind of continuity. 3. Chapter 24 is likely to formalize the citation-versus-connected comparison as one instance of the broader choice between phonemic, broad phonetic, and narrow phonetic transcription levels, asking learners to decide how much connected-speech detail a given transcription purpose actually requires.
Looking Ahead
What You Can Now Do
You can now distinguish coarticulation from categorical phonological alternation, identify assimilation, reduction, deletion, epenthesis, resyllabification, and linking in connected speech, compare citation forms with connected realizations, explain how style and rate affect the amount of connected-speech change, and produce a controlled connected-speech transcription using only previously taught notation. Together with Chapters 19 through 22, you now have a complete introductory picture of how speech is organized above the level of the individual segment, from the syllable through prosody to the changes that occur when words meet in fluent speech.