Please enable JavaScript.
Coggle requires JavaScript to display documents.
9) Word Recognition & Reading (Reading aloud (Connectionist triangle…
9) Word Recognition & Reading
Research + Research Methods
Reading
Unlike speech: children do not acquire reading w/o explicit instruction (has to be taught)
Culturally acquired skill
Largely automatic to skilled reader
Complex: perceptual, cogntiive & linguistic processes
Semantics: meaning of words
Syntax: rules for combining
Phonology: sound of words
Discourse processing: making inferences
Orthography: spelling of words
Research Methods
Reaction time tasks: speeded responses
Semantic categorisation task (semantic): is hawk an animal?
Naming pronunciation (read aloud: phenology): saying a printed word out loud as quickly as possible
Lexical decision task (orthography): is TRUCK a word?
Automaticity of reading
Stroop colour
Masked priming
Indirect study of word processing: studying influence of distractor or prime on target
Neuroimaging (ERP, MEG, fMRI
Eye movements during reading
Neuropsychology (study of patients)
Silent word recognition
Interactive activation model
(mcCelland & Rumelhart)
Single word recogntition
Problem with IA model: slot-coding
Cambridge email
First & last in correct position: rest jumbled
Should not be able to be read & comprehended
Cannot explain the transposed-letter priming effect
Transposed-letter (TL) priming effect
Response to target (JUDGE) is faster if it is primed by
jugde
(transposed letter) than by
junpe
(substituted letter)
Inconsistent with slot code assumption (letter position is coded precisely)
Masked priming task
Assumes letter position is coded precisely: cannot capture similarity between words in wrong order: CAT and ACT are just as dissimilar as CAT and BUT
Word superiority effect
Perceptual identification task: identifying letter is easier in a word context
Top down feedback: from word to letter level
Inhibition
within
the word level
About the model
Botton-up & top-down excitation/inhibition
Within-level inhibition: different features/letters/words are mutually inhibitory
3 levels of detectors: feature (\ / _), letter (A), word (CAT)
Basis of all models of reading (eg. DRC)
Automatic Processing
Retrieval of semantics
Semantic Stroop Effect
Semantic retrieval = automatic;
incongruent at level of meaning
Not reduced/eliminated by typical: presenting colour w/n single word, hypnotic word blindness suggestion
SKY (green) > PUT (green)
Semantic = controllable:
SKY-PUT effect is small & not robust: using colour words not in response set (eg. PURPLE) is larger and modulated by attention
Greater interference with (incongruent) colour-associated word distractor than colour-unrelated words
Retrieval of phonology
Homophone interference effect (van Orden, 1987)
Same pronunciation, diff spelling: ROWS/ROSE, EWE/YOU
More error to a non-member when it is a homophone of a category in semantic categorisation
Is it a flower: ROWS > ROBS (mandatory phonological processing: mistake ROWS for ROSE)
Masked phonological priming effect
Word processing = faster when proceeded by phonologically identical nonword primes > primes similar in orthography
klip-CLIP > plip-CLIP
Mask = prime not consciously percieved
Phonological processing = rapid & automatic
Is sound of word automatically retrieved in silent reading
Frost (1998)
Strong phenology hypothesis
Phonological processing = a necessary product of processing printed words
Phonological processing = mandatory in reading
Word identification
Fast
: speeded response task
Unavoidable
: Stroop colour naming task
Incongruence: Interference indicates word identification is obligatory
Unconscious
: masked priming; prime presented & masked so ss cannot identify
Compare effects of related prime vs. unrelated (control) prime
Orthographical: eagle-EAGLE
Semantic: hawk-EAGLE
Phonological: eegle-EAGLE
Reading aloud
Dual-route cascaded (DRC) model
Key assumptions
Non-lexical route:(1)
spelling-to-sound
Lexical route: (2 & 3)
uses lexicon, sensitive to frequency to useage (high freq: irreg = reg low: irreg > reg)
Connectionist triangle model
Two pathways when reading outloud
Direct: Orthography to phonology
Indirect: through word meaning
Semantic knowledge = largest impact on inconsistent words because of longer processing time
All information can be used to read words & pseudo words (orthography, phonology, semantics)
Irregularities resolved via cooperate & competitive interactions
Varieties of acquired dyslexias
: Due to brain damage; previously literate
Phonological dyslexia
Intact word reading: poor at reading nonwords eg. JINK
Patient WB
Couldn't produce sound of any given letter
Able to use lexical route, but not non-lexical (difficulty with grapheme-phoneme)
Representation of familiar words are stored in lexicon: meaning is activated; sound pattern is generated in the phonological output lexicon
Deep dyslexia
Phonological dyslexia + semantic errors: reads TULIP as ROSE
Difficult reading absract & function words: "the/and"
Recovering: become phonological dyslexics
Outside scope of DRC model
Resembles a severe form of phonological dyslexia
Surface dyslexia
Intact non-word reading, poor at reading irregular words
Patient KT
Regularisations: PINT --> mint, YATCH ---> hatched
100% nonword, 81% regular, 41% irregular
Over-reliance on non-lexical route: mapping spelling (graphemes) onto sound (phonemes)
Can be explained in terms of damaged to semantic system: strong association between impaired semantic knowledge and surface dyslexia in semantic dementia patients: if semantic knowledge is intact; can read irregular words
Alphabetic writing systems:
phonemes
(smallest unit of sound) represented with
graphemes
(visual letter/letter cluster)
Orthographic depth
Deep/opaque
Spelling to sound mappings are ambiguous/inconsistent
OUGH: THOUGH, TOUGH, COUGH
((PINT)), MINT: irregularity
Reading development slower in English
Shallow/transparent
Grapheme to phoneme is predictable/consistent
Italian, Spanish, Finnish, Czech
Words VS non-words
PINT (irregular word) VS JINK (pesudoword)
JINK: read using spelling to sound mapping rules
PINT: violates rule: must have a stored pronunciation (lexicon = mental dictionary)
Eye movements in reading
Saccades & fixations
Reading & eye movements
Fixations
: 250ms
Regressions
: infrequent right to left movements 2-5 characters
Saccades: 20-30ms:
move around 7-8 characters: "see nothing at all"
End of line sweep
single long saccade from end of one line to beg next
Eye tracking
Watch eyes move
Control what you see and when
Present sentences on screen
Saccades
20-30ms
8 seconds duration
Ballistic: initiated & direction cannot change
Seperated by fixations 200-250s
Rapid jerks
Information is not extracted: only during fixations
Effective FOV: 3-4 left, 15 to right
Fixations
20& function words: articles, conjunctions, pronouns
Fixation duration longer for "more difficult" words (rare, less predictable words)
80% of content words: nouns, verbs, adjectives
Word not fixated: common, short, predictable
Application: Spritz
Claims
Eliminates time scanned for n