Reading Flashcards
LANGUAGE
- “… man has an instinctive tendency to speak as we see in the babble of young kids while no kids have instinct to bake/brew/write…”
IMPORTANCE IN COGNITION
- written/spoken/nonverbal
- communicate our thoughts/ideas/feelings/needs
- skills allow us to put thoughts into words (cognitive) emotions (emotional) to then give encountered people (social)
- influences world perception
WORD DEFINITION
- form (sign) w/function (what's signified) FORM - pronunciation (phonology) - spelling pattern (orthography) FUNCTION - meaning (semantics) - syntactic role
SPOKEN WORD FORM
- phenomes (smaller sound unit) sequence
- organised into syllables
- w/stress pattern (ie. contEnt VS cOntent)
- tone/pitch pattern (some languages ie. Chinese; eg. ma (level/rising/dipping/falling) = mother/hemp/horse/scold)
WRITTEN WORD FORM
- graphemes sequence (lines/curves/strokes)
- dif language scripts use dif principles:
ALPHABETIC = graphemes represent phonemes (not always transparent); ie. Arabic/Thai/Hebrew/English
SYLLABIC = graphemes = syllables; ie. Korean/Kana
IDEOGRAPHIC/IOGOGRAPHIC = graphemes = meaning units; ie. Chinese/Kanji
COGNITIVE READING ANALYSIS
- skill performed for 5000+ years; invented in ancient Sumeria (Iraq); many still illiterate
- for most literate adults = familiar/automatic as breathing; impossible not to if viewing text
- complex despite ease/automaticity; multiple components
LEXICAL ACCESS
- components of reading = speech/text -> words; signal -> meaning
1. identify letters; represent sequence
2. identify words
3. retrieve syntactic class (other usage properties)/word meaning (concepts)
4. interpret sentence structure/meaning
5. interpret intention of speaker/writer (via extralinguistic context & knowledge)
SPELLING/PRONOUNCIATION/MEANING
- spelling/pro = relation varies across languages; semi-regular in English
- spelling/meaning = arbitrary relation
- pro/meaning = arbitrary relation
- SO reading requires process that identifies written form (ie. matched to learned memorised form); retrieves associated meaning
- if script = phonologically transparent, translation spelling/pro = done via rule w/o word identification
WORD IDENTIFICATION
- w know +100k word forms
- typical reader identities 2/3s
- presentation rates = 10p/s; some comprehension ok
EVIDENCE - introspective reports; observation/measurement/manipulation of beh; measurement/manipulation of brain activity
ANALYSIS LEVELS - experiential; computational/functional (processes/modules/architecture); neural (how neurons implement computations/where they are)
BEHAVIOURAL MEASURE
ARTIFICIAL
- lab tasks designed to exercise/capture component process
- typically discrete stimulus -> response tasks allowing accuracy/reaction time measured to each stimuli
ON-LINE
- measures made during continuous “natural” performance of skill
SINGLE WORD IDENTIFICATION LAB TASKS
- naming reaction time; non-essential word identification for “scombroid” BUT is for “pint”
- lexical decision (word/nonword) reaction time (dracknell/child)
- we can earn word recognition by getting people to identity individual letters in strings
WORD-SUPERIORITY EFFECT IN BRIEF EXPOSURES
REICHER (1969)
- task is to identify a briefly flashed letter
- performance better in word condition in spire of control for guessing (ie. d + k = word)
LAB PHENOMENA
WORD SUPERIORITY EFFECT
- greater accuracy of letter identification in word context than matched non-word
FREQUENCY EFFECTS
- RTs for lexical decision/sematic categorisation; naming shorter for words more frequent in language
SENTENCE CONTEXT EFFECTS
- RTs for lexical decision; naming shorter when word presented in sentence context of which its a plausible continuation
NORMAL READING EYE FIXATION DURATIONS
RAYNER & POLLATSEK (1988)
- fixation durations = 66-416 range; 218 msec
- saccade length = 1-18 range; 8.5 characters
- regressions = 10-15%
‘MOVING WINDOW’ TECHNIQUE
RAYNER & MCCONCKIE et al
- window extends 8 characters after + before fixation; reduction til reading performance suffers = shows how far ahead of/behind info is being taken in
- perceptual span (letters taken in) = 3-4 left letters; 14-15 right
WORD READING THUS FAR
- spelling pattern (orthography) -> meaning (semantics)
- arbitrary relationship; requires spelling pattern identification
- investigating lexical identification requires performance measure in tasks ie. lexical decision/categorisation/word naming, finding:
- higher frequency words recognised easier/faster
- letters easier recognised in word context
- words easier recognised in sentence context
- similar frequency/context effects for reading fixation durations
THEORIES OF WORD IDENTIFICATION
- finding best match between input-1000s spelling patterns in memory requires comparison process; one serial pattern/all parallel patterns? questioned
SERIAL SEARCH MODEL
MODIFIED SERIAL MODEL
PARALLEL-MATCHING WORD-DETECTOR MODEL
SERIAL SEARCH MODEL
FORSTER (1976)
- encode spelling pattern
- compare one at a time to each word-form stored in mental dictionary
- if match found, retrieve meaning/pronunciation; if not, continue search
- in order of language frequency for efficiency; match faster for higher-frequency words
- BUT simple; thousands of successive comparisons p/s; easy in computer, too fast in neurons
MODIFIED SERIAL MODEL
MURRAY & FORESTER (2004)
- mental lexicon of word-forms divided into “bins”
- quick/dirty initial process categorises spelling pattern to select correct bin
- serial search (frequency ordered) within bin
PARALLEL-MATCHING WORD-DETECTOR MODEL
MCCLELLAND & RUMELHART (1981)
- interactive activation model with detector units
- TIME -> feature units -> letter units -> word units = top-down activation of letter units by word units; explains effect of context
- accounting for word superiority effect; better recognition of a letter in word context
FORSTER’S SERIAL SEARCH MODEL
- frequent words quickly recognised as:
- motivated by its ready account of frequency effect; lexicon searched in order of frequency
- prediction = advance frequency knowledge; if helps, lexicon decision helped mostly in low frequency words
- search can skip high frequency part of list
IA-STYLE PARELLEL-MATCHING PROCESS
- frequent words quickly recognised as:
- most-used detectors most sensitive (ie. higher resting levels/stronger connections; faster to activate)
- prediction = advance knowledge of frequency should help decision most for high frequency words
ADVANCED FREQUENCY KNOWLEDGE X LEXICAL DECISION
GORDON (1983)
- high/medium/low frequency words given
- mixed (can’t predict frequency) VS separate blocks (can)
- knowing frequency in advance helps for high NOT low frequency words; consistent w/parallel NOT serial model
BRAIN EVIDENCE
PINKER (1997)
- “… the mind is what the brain does…”
- cognitive psychologists want to model what the mind (embodied in brain) does
- to inform models, make use of:
- brain-damaged patient performance
- non-invasive measures/manipulation of brain activation in performance
NEUROPSYCH X ARCHITECTURE OF READING
- from print to meaning via pronunciation/direct?
- long before we learn to read we recognise many spoken words and their meanings/produce words to express meaning; pathways include:
1. text -> orthography (spelling pattern); DIRECT -> semantics (meaning) -> phonology (pronunciation) -> output
2. text -> orthography (spelling pattern); PHONOLOGICALLY-MEDIATED ACCESS -> phonology (pronunciation) -> output
PHONOLOGICAL DYSLEXIA
- evidence that phonological mediation isn’t necessary aka. damage to phonological route (phono dyslexia)
- some brain-damaged patients can understand some written words but can’t access sound pattern:
ELLIS & SIN (1983) - patient RD; CHAOS -> kwost; people muddled = chost
LEVINE et al (1982) - patient EB; reading comprehension = slow but accurate; unable to choose which 2/4 words sounded/rhymed same
HOMOPHONE CATEGORISATION ERRORS
VAN ORDEN (1987)
- hare VS harp/hair; more false positive errors to lure that sound the same as category members than to visually similar control items
- why are more mistakes made for homophones when classifying by meaning?
- access meaning via phonology
ROUTE CONTRIBUTION TO NORMAL READING
- recall depends on familiarity w/word
- if both test word (ie. hair) and homophone (hare) = high frequency, homophone effect disappears
- high frequency words don’t produce much of homophone effect; low frequencies do
- if repeated experience = strong O -> S mapping (ie. high frequency word), spelling activates meaning fast enough for semantic decision/comprehension before indirect meaning activation via pronunciation; if not (ie. low frequency), phonological mediation contributes
TEXT/DISCOURSE COMPREHENSION END PRODUCT
- mental model/situational representation = rep of meaning conveyed/constructed in memory as we read (who is doing what/to what/whom/where/when) in IRL/imaginary world
- active newly formed rep in WM (aka. on-stage play)
- language/text NOT only vehicle for mental models (direct experience ie. movie also does this)
- model is NOT represented in language but propositional “language of thought” specifying elements/relationships (though linguistic tokens (ie. names) CAN be tagged)
- not just pictorial (though imagery represents some elements)
EARLY-LATER STAGES OF COMPREHENSION
identify words -> retrieve:
- SYNTACTIC class (other usage properties) -> interpret sentence structure (syntax) -> interpret/link constituent propositions -> interpret intention of speaker/writer -> MENTAL MODEL
- WORD meaning (concepts) -> interpret/link constituent propositions -> interpret intention of speaker/writer -> MENTAL MODEL
- (all combos always include mix of extralinguistic context/knowledge)
THE SPEECH ART
- sentence meaning = propositions stated + speech art (querying/stating/denying, etc.)
- sentences give explicit/implicit propositions to extract/link (ie. “my daughter phoned me from Australia” = X is Y’s mother; Y located in A at T (time); Y phoned X at T)
- understanding connected discourse requires linking given/new propositions within/across sentences ie. via co-reference (ie. “she was driving to Broome” = Y (at T, A) on way to A2 (Broome); Y driving (a vehicle))
- comprehension activates/adds propositions to existing knowledge in memory
BUILDING PROPOSITIONS
- sentences = tree structures/ordered hierarchy of constituents (phrases) occupying essential roles in relation to main verb (subject)
- constituents replaceable by others of same type w/o changing structure = innumerable possible sentences BUT limited structure range p/language
SENTENCE STRUCTURE CLUES
WORD ORDER
- bishop admired actress VS actress admired bishop
FUNCTION WORDS
- opposed to content words
- small fixed sets of grammar words structure-signalling ie.
- the/a/an = determiners; introduce noun phrases
- who/which/that = relative pronouns; introduce relative clauses
WORD-MODIFYING “MORPHOLOGICAL INFLECTIONS”
- not many in English
- signalling number/case/tense etc.
- dogs/dog/dog’s
- I talk/talked
- tall/taller/tallest
PARSING BRAIN MODULE EVIDENCE
- parsing = specialised structure-computing
- Broca’s aphasia patients struggle w/comprehending:
- syntactically complex sentences (ie. put small round green circle to left of brown square above triangle)
- simple reversible sentences (ie. pictures of dancer/clown applauding each other; selection)
- sentences dependent on affixes/function words for meaning
LEVELS OF LANGUAGE AMBIGUITY
LEXIAL AMBIGUITY - words w/several distinct meanings (ie. bank/fine) SYNTACTIC AMBIGUITY - ambiguous sentence structures - ie. "visiting relatives can be boring" SPEECH ACT AMBIGUITY - pilot: "We are now at take-off" - tower: "OK" (permission?) AMBIGUITY OF REFERENCE - ie. John pointed out Bill's friend, Paul, who gave HIM HIS hat
INTERPRETNG WRITE/SPEAKER INTENTIONS
- utterances have surface forms = directly indicate speech act (ie. declaration/question etc)
- any form uttered in suitable context w/particular intonation can perform indirect speech act (ie. “It’s cold in here” + looking meaningfully at door = close it request)
- intended speech act must be inferred via prosody/extralinguistic cues (gaze/body language)/context/general knowledge
- social/legal acts signalled via particular performatives (phrases) but work only in felicitous context (ie. I declare thee man and wife)
INFERENCE X COMPREHENSION
- discourse explicitly states only some propositions needed for coherent mental model construction; infer rest based on:
- extralinguistic context (immediate environment/assumed status/intentions of speaker/writer)
- prosody/body language (gesture/expression)
- linguistic context (what’s been said already)
- general knowledge (properties/characteristics/schemas)
- communication conventions (relevance)
INFERENCE EXAMPLES
CONSEQUENCE
- actress fell from the window on her 14th floor flat onto piazza = death inference
INSTRUMENT
- she was driving to Broom = vehicle instrument inference
SPACE/TIME
- Alex went to lecture; found dead frog under seat = time/location of action inference
CAUSE
- students carelessly threw cig; fire destroyed campus = students were cause of fire inference
AUTOMATIC INFERENCING
GARNHAM (1979)
- tested cued verbatim recall for sentence lists (ie. John cooked the chips)
- FRY > COOK retrieval
- we infer automatically and remember inferences as if explicitly stated even when instructed to trust verbatim
SOLVING LEXICAL AMBIGUITIES
- many words/structures = ambiguous till later info instructs interpretation (ie. he went to bank TO cash cheque/catch fish)
- usually disambiguate meaning unconscious of ambiguity/noticeable perturbation BUT average fixation durations = longer on ambiguous words
- sometimes have to backtrack to make sense of lexical ambiguity (ie. rapid righting/writing of the canoe saved him) or syntactic ambiguity (ie. the horse raced past the barn fell)
- garden path sentences in reading = regression eye moves
LOCAL AMBIGUITY STRATEGIES
MINIMAL COMMITMENT STRATEGY
- postpone interpretation until all potentially disambiguating info available = NOT PLAUSIBLE! priming recognition via incomplete contexts indicates incremental comprehension
SERIAL STRATEGY
- construct most probably interpretation; back track if wrong
PARALLEL STRATEGY
- construct multiple interpretations; delete the wrong ones
ALL REQUIRE WM!
SEMANTIC PRIMING EFFECT
MEYER & SCHVANENELDT (1971)
- index of activation of meaning
- provides measure of activation of meaning of prime
- use to ask if both meanings are ambiguous word activates
SWINNEY et al (1979)
- “government plagued… man was not surprised to find spiders, roaches and other bugs in the corner of his room…”
- ant = appropriate; -38ms/-53m RT immediately/after
- spy = inappropriate; -31ms/-1ms RT immediately/after
- parallel activation of meanings = inappropriate meaning suppressed
FIXATION DURATIONS X LEXICAL AMBIGUITY
RAYNER et al
- if no disambiguating context before + equal frequency meanings = ambiguous > unambiguous fixations
- if disambiguating context before ambiguous word = ambiguous = unambiguous fixations (both meanings un-activated)
- unless contextually appropriate meaning lower in frequency = higher frequency meaning gets activated even if not supported by prior context
SUMMARY
- meanings of ambiguous word activated in parallel BUT not w/equal strength
- relative strength of activation depends on:
- degree of contextual constraint available
- frequency of use of each meaning
