Word recognition Flashcards
What did Moon et al (2013) find about babies hearing and sucking dummies?
- 30-hour old babies were exposed to vowels unique to their mother’s language (English or Swedish) and to vowels unique to the other language.
- Babies sucked their dummy harder when listening to the foreign language – not already familiar with
- Hearing system is mature enough 7 months after conception, babies can learn about speech
- Suggests that babies listen to speech in the womb and are learning about it!
Which two characteristics of speech make word recognition difficult?
- The segmentation problem (ice cream vs I scream)
- The invariance problem
What is the segmentation problem?
- Speech is a continuous stream of sound. No white spaces between words
What are the solutions to the segmentation problem?
- Possible word-constraint (Norris et al., 1997), “fapple” vs. “wuffapple” - any segmentation that results in impossible words is likely rejected (e.g. “fapple”), segmented into wuff sound and apple
- Stress (Cutler & Butterfield, 1992) - first syllable typically stressed, “conduct ascents uphill” heard as “the doctor sends the bill” (stressed syllable underlined)
- Different languages have different strategies
What is the Invariance problem?
- Speech sounds (e.g. phonemes) are shaped by external factors.
- Co-articulation - /b/ in “bill”, “ball”, “bull”, “bell” is acoustically different
- Bacon example: lean, bacon, lean bacon (how tongue and lips move)
- Has to take in variability, makes speech more fluent
- Dialect, accent, speech rate…
What is the - Phoneme restoration effect (Warren & Warren, 1970) ?
– cough sound replacing a phoneme harder to detect what word is being affected compared to silencing a phoneme
- The state governors met with their respective legislature convening in the capital city (sentence example)
- Mind constantly filling in missing gaps in speech
What was the *eel test?
- It was found that the *eel was on the orange (peel)
- It was found that the *eel was on the axle (wheel)
- It was found that the *eel was on the shoe (heel)
- It was found that the *eel was on the table (meal)
- Cough in front of eel changing how word is heard
Two of the most prominent theories in spoken language are …
TRACE and the Cohort model
What is TRACE (McClelland and Elman, 1986)?
- Word units (e.g. cat, bad, bat)
- Phoneme units (activated at this level, starts to inhibit competing phonemes. Inhibition makes it easier for one phoneme to be the dominant one)
- Feature units
- Information can flow bottom-up and top-down across the three levels!
What does TRACE explain?
- TRACE can explain context effects
- This is because it allows higher-level information to affect lower-level information (top-down effects)
What does TRACE overestimate and what’s an example of this?
- TRACE overestimates the influence of context and predicts top-down effects where they do not exist
- Frauenfelder et al. (1990): TRACE predicts it’s hard to detect /t/ in (French) nonword vocabutaire because of the similar real word vocabulaire. Data did not support prediction.
- French participants heard French words and non-words
What is the Cohort model - Marslen-Wilson (1984)?
- Speech unfolds overtime compared to written language
- table tea tide tomorrow tooth traffic trainer and all other words that begin with the /t/ phoneme!
- When more information/phonemes of the word come in you can eliminate words which don’t match the cohort
- Remove competitors, allows you to eliminate all but one candidate
In the original Cohort model, the selection stage is influenced by…
- In the original Cohort model, the selection stage is influenced by the auditory presentation of the word (like seen in previous slides) and the semantic or syntactic context
- A word can be recognised before its uniqueness point if context supports only one candidate in the cohort
- Example: “The poacher ignored the sign not to tres-” … In this sentence context “trespass” has a recognition point that comes before the uniqueness point
The original Cohort model failed to deal with two problems, which are?
- If the first phoneme is mispronounced or misperceived, the word should never be recognised
- but hearing focabulaire activates vocabulaire (Frauenfelder et al., 2001)
-Context does not eliminate words from the cohort! What happened in Zwitserlood(1989) semantic priming experiment?
- E.g. the word ‘cap’ being primed to ship and money
- The part-word cap… semantically primes both ship and money. So both captain and capital were activated when hearing cap….
- Another example: “With dampened spirits the men stood around the grave. They mourned the loss of their cap…”. Original Cohort model says capital should be eliminated from the cohort by the sentence context. But it was not! Cap.. still primes money!
