2. Spoken Word Recognition

Question 1

What are the main reasons that speech differs to text?

Answer 1

1. Variability in signal
2. Distributed in time
3. No gaps between words
4. Phonological context effects

Question 2

Why is there variability in speech signals and how do we overcome this?

Answer 2

• Place of articulation, voicing vary
• No sharp boundaries between similar phonemes
• differences between and within speakers

Categorical perception- reduce variability of Signal
(Eimas et al., 1971)
• Boundaries of enhanced discriminability
• Different languages have different boundaries

Lexical knowledge
(Ganong, 1980)
• ambiguous phonemes inserted into context- participants hear word
(Warren, 1970)
• Phoneme restoration effect- one sound in sentence replaced with cough

Specific speaker knowledge
(Nygaard and Pisoni, 1998)
• better identification of words given by familiar speaker

Visual information
McGurk effect- listeners make use of speakers lip position

Question 3

How does the fact speech is distributed in time affect recognition?

Answer 3

Whole word not immediately available

(Zwisterlood, 1989)
Cross-modal priming evidence for early recognition
• Hear ambiguous fragment, then lexical decision to visual target
• capt primes both ship and guard, captain only ship
—> listeners initially activate both

(Marlsen-Wilson,1987) and others
• Cohort view of word recognition

Question 4

How does having no gaps between words make speech differ to text?

Answer 4

• Segmentation- acoustic cues, plus words recognised before offset so listener knows next onset

(Bard et al., 1988)
• up to 60% words not unique before offset

(Cutler, 1990)
• Metrical Segmentation Strategy MSS
• most words start with stressed syllable

(Cutler, 1990)
• Possible word constraint PWC
• Prefer to segment speech so every sound is accounted for by a possible word

—-evidence: (Norris et al., 1997)
• Words harder to detect when preceding sound can’t be a word on its own

Question 5

How do phonological context effects mean speech differs to text?

Answer 5

Phonological context effects- speech sounds influenced by adjacent sounds
• Coarticulation- smearing
• Place assimilation- place of articulation can move to make speech easier

(Warren and Marlsen-Wilson, 1988)
Scoot and scoop vowels provide info about which word is coming