Questions: Coarticulation and Phonetic Context Effects in Speech

Question 1

A researcher excises a /d/ token recorded from the word 'dim' (where /d/ is followed by a high-front vowel) and splices it before 'oom.' Listeners still report hearing /d/. What does this best demonstrate?

• A. The acoustic signal for /d/ is consistent regardless of surrounding vowel context, so the token sounds like /d/ in any setting
• B. The perceptual system compensates for coarticulatory context, recovering the intended phoneme from a signal shaped by a different phonetic environment
• C. Categorical perception forces a binary consonant choice regardless of the specific acoustic evidence
• D. Listeners rely on lip reading rather than acoustic signals to identify consonants

Answer: (B)

This is the perceptual compensation phenomenon. The acoustic token was shaped by high-front vowel context (formant transitions appropriate for /iː/), yet listeners identify /d/ — the intended segment — when it occurs before a different vowel. This shows the perceptual system does not passively match acoustic templates; it infers the underlying phonological gesture by factoring out the coarticulatory context. Option A is the central misconception: if consonant acoustics were context-independent, coarticulation would not create a variation-invariance problem in the first place.

Question 2

When saying 'stew,' a speaker's lips begin rounding before the /s/ and /t/ consonants are finished. This is best described as:

• A. Carryover coarticulation — articulatory state from a preceding phoneme persisting into the current one
• B. Anticipatory coarticulation — the gesture for an upcoming phoneme already shaping current articulation
• C. A production error caused by insufficient articulatory planning
• D. Evidence that phonemes are produced as discrete, non-overlapping units in a strict sequence

Answer: (B)

Anticipatory coarticulation occurs when an upcoming segment's articulatory requirements begin influencing current production. Lip rounding starting during /st/ in anticipation of the /uː/ vowel is a classic example — the vocal tract is already preparing for a segment that hasn't started yet. Carryover coarticulation (option A) is the reverse: a preceding phoneme's articulatory state persisting into the following one. Option D represents the misconception that coarticulation is designed to refute: speech is not a sequential chain of discrete, non-overlapping gestures.

Question 3

The same phoneme produces acoustically consistent signals across different phonetic contexts, and this invariance is what allows listeners to reliably identify phoneme categories.

• T. True
• F. False

Answer: False

This is precisely what coarticulation disproves. The same phoneme produces systematically different acoustic signals depending on what phonemes surround it — the /d/ in 'deed' has different formant transitions than the /d/ in 'dood' because the surrounding vowels shape the tongue position during the consonant. Acoustic variability is the rule, not the exception. The variation-invariance problem is the challenge of explaining how listeners achieve stable perceptual categories despite this variability — not by denying that variability exists.

Question 4

Coarticulation means that articulatory gestures for neighboring phonemes overlap in time, so the same phoneme is physically produced differently depending on its phonetic context.

• T. True
• F. False

Answer: True

This is the core definition of coarticulation. Because articulators move continuously rather than teleporting between positions, the gesture for an upcoming phoneme can begin before the current one ends (anticipatory coarticulation), and the state from a preceding phoneme can persist into the following one (carryover coarticulation). This temporal overlap means the same phoneme — say, /d/ — is realized with different articulatory specifics (and different acoustic outputs) in 'deed' versus 'dood.' Understanding this is key to understanding why the variation-invariance problem exists.

Question 5

Why does coarticulation evidence support the view that speech perception is inferential rather than passive acoustic analysis?

Model answer: Because the same phoneme produces systematically different acoustic signals in different contexts, yet listeners always identify it correctly. If perception were passive acoustic pattern-matching against stored templates, this variability should cause frequent errors. Instead, listeners correctly identify phonemes even when the acoustic token was shaped by a different phonetic context — demonstrating that the perceptual system actively infers the intended phonological gesture by compensating for the surrounding phonetic environment.

The key logical move is from variability of input to stability of output. Passive analysis predicts errors when the acoustic signal deviates from the stored template for a phoneme. Perceptual compensation experiments show that listeners are robust to exactly this kind of acoustic variation when it is explainable by coarticulation. The system is not asking 'what acoustic pattern do I hear?' but rather 'given this acoustic signal and this phonetic context, what phoneme was the speaker intending to produce?' — an inferential process that uses knowledge of how sounds are produced, not just how they sound.