Language and reading Flashcards
Lectures 15+16 (44 cards)
1
Q
importance of language in cognition
A
- language can be written, spoken or nonverbal
- through it we communicate our thoughts, ideas, feelings and needs
- language skills allow us to put our thoughts into words, words to our emotions and to communicate these thoughts and emotions to the people we encounter daily
- language can influence how we perceive the world
2
Q
what’s a word
A
- a form with a function
- a pronunciation with a meaning
- a spelling pattern with a syntactic role
3
Q
spoken word form
A
- a sequence of phonemes
- organised into syllables
- with a stress pattern
- and in some languages a tone
4
Q
written word form
A
- a sequence of symbols made of lines, curves or strokes
- different languages’ scripts use different principles
- alphabetic - graphemes represent phonemes
- syllabic - graphemes represent syllables
- ideographic/logographic - graphemes represent meaning units
5
Q
reading - analysis of a cognitive skill
A
- skill performed for about 5000 years, since writing was invented
- many people in the world still don’t do it, but for most literate adults reading is a task that is almost as familiar and automatic as breathing
- impossible not to read if you are looking at text
- reading is a complex skill made up of many components, in spite of its ease and automaticity once you can do it
6
Q
components of reading
A
- lexical recognition - recognising words
- syntax and semantics - understanding how words fit together
- ambiguity in language processing - how vague or complex instructions require interpretation
7
Q
component processes in reading for meaning
A
- after retrieving words and their meanings, brain processes sentence structure, sentence meaning, and intention of speaker/writer
- extralinguistic context and knowledge - understanding meaning also involves prior knowledge, cultural understanding, communication context
- reading comprehension involves multiple layers of processing, from identifying letters and words to understanding sentence structure and meaning
- extralinguistic factors help in fully interpreting a sentence
8
Q
translation between spelling, pronunciation and meaning
A
- relationship between spelling, pronunciation and meaning is arbitrary - words must be learnt not deduced
- different languages use different word forms for the same meaning
- reading involves recognising and retrieving word meanings, with some languages making this easier through phonological transparency`
9
Q
word identification - challenge
A
word identification - challenge
10
Q
sources of evidence
A
- introspective reports
- observation, measurement and manipulation of behaviour
- measurement and manipulation of brain activity
11
Q
level of analysis
A
- experiential
- computational, functional
- neural
12
Q
two kinds of behavioural measure
A
- artificial lab tasks designed to exercise and capture a component process, typically discrete stimulus response tasks, allowing accuracy and/or reaction time to be measured to each stimulus
- on line measures made during continuous natural performance of the skill
13
Q
lab task for probing single word identification
A
- naming reaction time, word identification not essential for real word but it for nonsense word
- lexical decision reaction time
- can also learn something about word recognition by getting people to identify individual letters in letter strings
14
Q
word superiority effect in brief exposures
A
- task is to identify a briefly flashed letter
- performance better in word condition in spite of control for guessing
- top down processing - when a letter appears within a word, the brain uses prior knowledge of familiar words to aid recognition
- contextual facilitation - words provide contextual clues that help reinforce letter identification
- parallel processing - the brain processes entire words, rather than individual letters, allowing for more efficient recognition
15
Q
some phenomena observed in these lab experiments
A
- word superiority effect - greater accuracy of letter identification in the context of a word than of a matched non-word
- frequency effects - RTs for lexical decision, semantic categorisation and naming shorter for words that are more frequent in the language
- sentence context effects - RTs for lexical decision and naming are shorter when a word is presented in a sentence context of which it is a plausible continuation
16
Q
eye fixation durations normal readings
A
- saccades are the eye movements between fixations - longer saccades indicated fluent reading
- short high frequency words tend to be skipped or recieve short fixations
- longer or less frequent words receive longer fixations
- regressions suggest comprehension difficulties or ambiguity
- eye movements are not smooth but jump from one fixation to another
- this research helps us understand reading efficiency and how our brain processes text
- findings are applied in areas such as dyslexia research, speed reading techniques, user interface design, eye tracking applications
17
Q
theories of word identification
A
- finding the best match between input and one of the 1000s of spelling patterns we have in memory requires a comparison process
- do we compare the input to one pattern in memory at a time or to all the patterns in memory at once
18
Q
serial search model
A
- encode spelling pattern
- compare one at a time to each word-form stored in the mental dictionary
- if match found, retrieve meaning and/or pronunciation, if not, continue search
- for efficiency, the search is in order of frequency in the language - hence a match will be found faster for a higher-frequency word
19
Q
modified serial model
A
- simple serial model requires many thousands of successive comparisons per second
- easy in a modern computer, much too fast for neurons
- Forster and Murray proposed - mental lexicon of word-forms is divided into bind, a quick and dirty initial process categorises spelling pattern to select appropriate bin, serial search within that bin
20
Q
interactive activation model
A
- reading is not a simple left-to-right process, multiple words get activated in parallel
- similar looking words compete before one is selected
- explains reading errors
21
Q
accounting for the word superiority effect
A
top down activation of letter units by word units can explain this effect of context
22
Q
why are frequent words recognised more quickly
A
- Forster’s serial search model - motivated by its ready account of frequency effect, we search the lexicon in order of frequency - prediction is that if this does help, it should help lexical decision for low frequency words the most
- IA style parallel matching process - can also account for frequency, most used detectors are the most sensitive - prediction is that this would help the most for high frequency words
23
Q
prediction tested - effect of advance knowledge of frequency on lexical decision
A
- high, medium and low frequency words presented - mixed vs separate
- knowing the frequency in advance helps for high but not low, which is consistent with parallel not serial model
24
Q
using neuropsychology to ask about architecture of reading
A
- long before we learn to read, we can recognise many spoken words, and retrieve their meaning of produce those words to express meaning
- direct pathway - orthography - semantics, reader can map written words directly to meaning without involving pronunciation, more efficient for skilled readers and common for high frequency words
- phonologically mediated pathway - orthography - phonology - semantics, readers first convert text into phonological representation before accessing meaning, more common for less familiar words and phonetic languages
- skilled readers rely more on the direct pathway
25
how do we measure reading speed
1. calculate words per minute (total words read / time taken)
2. 5-10 comprehension questions
26
evidence that phonological mediation is not necessary
1. some brain damaged patients can understand some written word but cannot access their sound pattern
2. supports the dual route model of reading - phonological route (damaged in these patients) is not always necessary
3. direct orthographic-to-semantic route allows reading comprehension without phonology
4. challenges serial phonological models, which assume all words must be sounded out before understanding
5. provides neuropsychological evidence that phonology is not the only route to meaning
6. helps explain phonological dyslexia, where patients can but read but struggle with pronunciation and sound based tasks
7. supports different reading strategies in individuals with reading difficulties
27
evidence that phonological mediation nevertheless occurs
1. Van orden - many more false positive errors to lures that sound the same as category members than to visually similar control items
2. we make more mistake for homophones when classifying by meaning because we access meaning via phonology
28
how do both routes contribute to normal reading
1. to depend on how familiar the word is
2. so if both test word and homophone are of high frequency, the homophone effect disappears
3. high frequency words don't produce much of a homophone effect, low frequency words do
4. conclusion - if repeated experience has established a strong 0-s mapping, spelling activates meaning quickly enough for semantic decision before indirect activation of meaning via pronunciation, if not phonological mediation contributes
29
end product of text comprehension
1. a mental model or situational representation - a representation of the meaning conveyed, constructed in memory as we read, who is doing what and to what
2. the active, just being formed representation is in working memory
3. language/text is not only the vehicle for creating a mental model
4. the mental model is not represented in language, but in a propositional "language of thought" or "mentalese" specifying the elements and their relationships
5. not just a pictorial image in the mind
30
from early to later stages of comprehension
1. role of extralinguistic context and knowledge: background knowledge fills gaps in understanding, cultural, contextual and experiential information enhance interpretation
2. comprehension happens in stage from basic word recognition to high level interpretation
3. understanding sentence structure and meaning is necessary before inferring speaker intent
4. extralinguistic knowledge helps form a coherent mental model of the text
31
sentence meaning
1. sentences state (explicitly and implicitly) propositions, which we must extract and link
2. to understand connected speech/text, we must link given and new information within and across sentences
3. the role of the speech act - the same words can have different meaning depending on the speech act
32
sentence structure clues
1. word order
2. function words
3. word modifying morphological inflections
33
syntactic structure
1. provides one kind of clue to the meaning
2. does not provide meaning
34
evidence for specialised structure-computing module in our brains
some Broca's aphasia patients have trouble comprehending syntactically complex sentences, simple reversible sentences and sentences whose meaning depends critically on affixes and function words
35
levels of ambiguity in language
1. words with several distinct meanings - lexical ambiguity
2. ambiguous sentence structures - syntactic ambiguity
3. ambiguity of reference
4. speech act ambiguity
35
to comprehend text
1. to interpret the speakers/writers intention - utterances have surface form which directly indicate a speech act: direct speech act (intention revealed by literal meaning), indirect speech act (literal meaning is not intended one), irony (literal meaning is not the intended one)
2. filling gaps with inferences - go beyond the literal meaning of text, we have to infer the unspoken based on the extra-linguistic context, prosody and body language, linguistic context, general knowledge, properties of things, communication conventions
36
some inference is automatic
1. Garnham - tested cued verbatim recall for lists of sentences, John cooked the chips but fry was a better retrieval cue
2. we infer automatically and remember what we infer as if it were explicitly stated
3. some inferences can be misleading - Loftus and Palmer
37
lexical and syntactic ambiguities
1. many words or structures are ambiguous until later information tells us how to interpret them
2. we usually dissolve the ambiguity without awareness or noticeable perturbation, but average eye fixation durations are longer on ambiguous words
3. we sometimes have to back track to make sense of an ambiguity
38
how do we solve ambiguities
1. postpone interpretation until all potentially disambiguating information available - minimal commitment strategy, not plausible as priming of words recognition by incomplete sentence contexts indicates comprehension is incremental
2. construct most probably interpretation, backtrack if it turns out to be the wrong one - serial strategy
3. construct multiple interpretations in parallel, delete those that don't work out - parallel strategy
4. all require working memory to hold on the input and represent the output
39
semantic priming
1. participants recognise butter faster when bread is the prime than when brain is the prime
2. this happens because bread and butter are semantically related making butter preactivated in memory
3. semantic priming shows that words are linked in out mental lexicon (word network)
4. it provides a tool for studying meaning activation in language processing, helps explore ambiguous words
40
using semantic priming as an "on-line" measure of the activation of the meaning of an ambiguous word
1. after hearing bugs, participants had to make a lexical decision about a visual probe word appearing on a screen
2. RT measured to see how fast they recognised words related to both meanings
3. immediately after hearing - faster RTs for both ant and spy in relation to bugs, suggests parallel activation
4. few syllables later - RT for ant remains fast, spy returns to baseline suggesting its suppressed
5. demonstrates how real time language comprehension works
6. shows that context does not immediately select the correct meaning, delayed
7. helps explain processing difficulties in dyslexia or second language learners
8. supports models of lexical access in psycholinguistics
41
fixation durations as an online measure of the processing cost of lexical ambiguity
1. if no diambiguating context before, and meanings of equal frequency, linger fixations on ambiguous word
2. if disambiguating context before the ambiguous word, fixations on ambiguous word no longer than for matched unambiguous, both meanings not activated unless the contextually appropriate meaning is much lower in frequency
3. looks as if the higher frequency meaning gets activated here even when not supported by the prior context
42
current conclusion on lexical ambiguity resolution
1. meanings of an ambiguous word activated in parallel but not with equal strength
2. relative strength of activation depends on degree of contextual constraint available, frequency of use of each meaning
43
language comprehension conclusions
1. comprehension is a constructive, integrative process that draws on background knowledge and results in the construction of mental models
2. inference is an important part of the construction process, it draws on knowledge schemas and scripts (knowledge of routine actions/sequences) in some cases inferences can be misleading
