Week Three - Speech Perception & Production Flashcards
What is the segmentation problem?
No clear boundaries between words
How do we solve the segmentation problem?
Possible word constraint: We like to segment speech so that it maps onto whole, possible words
Meaning Constraint: We prefer the mapping to make sense
What is the Invariance problem?
Phonemes are not always pronounced/perceived the same way
In what ways do phonemes vary?
Surrounding sounds
speech sound
speaker accent
speech formality
2 ways invariance problem presents itself?
Assimilation: Sounds take on some of their neighbours properties (song, Tom)
Coarticulation effects: Sounds can be produced more quickly/easily
What are allophones?
Phonemes pronounced slightly differently, but don’t contribute to differences in meaning (Tom, burton)
Voice onset time (VOT)
time between the burst and onset of voice (helps determine diff between eg p and b)
How does context help in identifying sounds/words?
Cause it is influenced by higher-level knowledge of semantics and syntax
When are single words better recognised?
In sentences than alone and against noisy background, and when same words are excised from sentences
Where does phonemic restoration occur?
At the word level
Template Matching Model of SP?
Target words are stored as templates. When we hear a word, it is matched to the stored mental template, recognition occurs.
Analysis by Synthesis Model of SP?
Motor theory: we interpret speech with reference to our own motor speech movements
Strength & weaknesses of ABS Model?
accounts for speaker differences
speech recognition seems driven by data and not hypothesis
Cohort Model of SP?
When we hear speech, we set up a cohort of possible words to decide which we heard. Items eliminated until one is left and this is assumed to be the word heard.
Stages of Cohort Model?
Access stage: representation of word
Selection stage: one item chosen
Integration stage: syntactic and semantic properties are used to integrate into sentence
Trace Model of SP?
Role of top-down processing, word and sentence context can facilitate perception of individuals sounds.
Nodes exist (interconnected) and activation matches the input stimulus
Vowels and consonants and air?
Vowels do not Obstruct the passage of air, consonants do
Pure vowel
Just one vowel sound
bat, bomb
Diphthongs
two vowels produced in a smooth glide that moves from one vowel to the other, bake
Three distinctive features of vowels?
tongue height, tongue position, lip position
Place of articulation
where sounds are made
Manner of articulation
how sounds are made
Rate of speech errors
1-2 per 1000 words
Two ways to collect speech error data?
- recorded corpora
- experiments
- elicit speech errors
Anticipation errors
letter sounds come earlier than intended (clounsil is getting closer)
Perseveration errors
sounds produced earlier reappear later on (eg annotated babliography)
Exchange errors
units of varying sizes change place (shrine spinks)
Importance of understanding speech errors?
Occurs at levels, suggests that each of these levels of unit is psychologically real and we must store and manipulate units of speech at diff levels
Garrett’s model of speech production?
- conceptualisation: pre-verbal message (ie i want to get a flat white)
- formulation: translated into linguistic form
- execution: articulation
Garretts stages of syntactic planning?
Functional
- content words selected
- meaning is specified
- words assigned to syntactic roles
Positional
- function words are selected
- words put into correct order
Word-blend error
two words must be simultaneously retrieved from the lexicon (thats what wuthers me) - bothers and worries
Phrase-blend error
phrases simultaneously activated and cross over where they are most alike (eg im making the kettle on)
Non-plan internal error
Message level info intrudes on lower levels of processing esp if sections sound similar
Environmental contamination?
Message level info intrudes on lower levels of processing esp if sections sound similar (context plays part) eg man running (its running outside)
