Mind map
LANGUAGE PROCESSING
Map content
LANGUAGE PROCESSING
SPEECH SEGMENTATION
- Pre-lexical
- Use characteristics of speech stream eg rhythm of lang (strong syllable in Eng)
- Can't account for all words eg confess
- CUTLER & NORRIS - wordspotting showed use of strong syllables to segment. Across cultures
- Use characteristics of speech stream eg rhythm of lang (strong syllable in Eng)
- Lexical
- Knowledge of how words sound, phonological representation
- Time span of short words doesn't allow for recognition
- SAFFRAN et al - babies (8 mths) using statistical occurrence of syllables. Turned to 'novel' speech.
- Developmentally linked - implicit learning ability
- Knowledge of how words sound, phonological representation
- Pre-lexical
MODELS OF SPOKEN WORD RECOGNITION
- PARALLEL ACTIVATION
- The cohort model - MARSLEN-WILSON et al
- Beginning of word heard
- word-initial cohort activated
- reduced as more sounds heard
- uniqueness point found - word deciphered
- reduced as more sounds heard
- word-initial cohort activated
- MARSLEN-WILSON - cross modal priming showed semantically similar words activated before uniqueness point
- Beginning of word heard
- Sounds evaluated and re-evaluated against possible matches
- Doesn't account for goal of word recognition
- The cohort model - MARSLEN-WILSON et al
- LEXICAL COMPETITION
- activation and competition
- can change due to info coming in - reflect strength of hypothesis
- McCLELLAND & ELMAN - TRACE connectionist model
- Phonetic level
- Phoneme level
- Word level + inhibitory links
- Phoneme level
- Nodes feed into each other with increasing activation
- McQUEEN et al - wordspotting embedded in sentences. Slower rec where syllables matched another word
- Phonetic level
- PARALLEL ACTIVATION
MODEL OF VISUAL WORD RECOGNITION
- IAC (interactive, activation and competition)
- McCLELLAND & RUMELHART
- Visual features
- Letters
- Words
- Letters
- Visual features
- Inhibitory nodes between levels - competition
- Lexical effects - word superiority effect (WSE)
- Letter detection easier when part of a word
- Word nodes provide secondary source of activation (top down)
- GRAINGER & JACOBS - WSE not necessarily top down as detection could be activation of word or letter nodes
- Word nodes provide secondary source of activation (top down)
- Letter detection easier when part of a word
- McCLELLAND & RUMELHART
- DUAL ROUTE - DRC
- COLTHEART et al
- Incorporates 'rule based' route to pronunciation
- AND
- Lexical route - retrieval of stored pronounciation
- AND
- Regular words make use of both routes - recognised quicker
- Irregular words - lexical route only - slower processing
- Cannot explain consistency effects
- GLUSKO - similar spellings effect processing. Regular words with diff pronunciations slower (wave, have, slave)
- SINGLE CONNECTIONIST MODEL
- SEIDENBERG & McCLELLAND
- Includes ir/regular words
- IAC (interactive, activation and competition)
EYE-TRACKING
- can indicate processing difficulties
- Show where ambiguity
- O-REGAN & JACOBS - optimal viewing position quicker word recognition
- SHILLCOCK et al - left favoured as more info to help process
Information
- Description
- Mind Map on LANGUAGE PROCESSING, created by knott76 on 02/10/2015.
- Created by
- knott76
- Views
- 3
- Rating
More on Speech Perception and Language Comprehension