Flashcards
Language and the Brain
Cards in this deck (71)
- 1competence vs performance?competence: idealized capacity/knowledge, linguistic theories (phonological, syntactic, semantic)
performance: dynamical, actual speech, how do we use this knowledge (=competence); contains slips of the tongue and false starts, represents only a small sample of possible utterances: "I own two-thirds of an emu" is a good English sentence (we have competence about it), but is unlikely to really be performed
- distinction because not all we produce is correct and we don't understand everything that is correct - 2phonological and lexical knowledge?- study of the sound system, what sounds can be produced, physical properties of sounds
- lexicon: vocabulary, consists of lexemes, has grammatical and lexical words, organized by fields but also events - 3syntactic knowledge?- what are the rules of how are words in the language combined? = grammar
- universal grammar (chomsky)?
- syntactic categories: noun, verb, preposition.. (this classification is based on similar structure and same relationship to other classes)
- - 4semantic/conceptual knowledge?- meaning of overarching concepts (concept chair) and words (stool, chair, armchair..)
- semantic triangle: pretty arbitrary relationships between names of things and things themselves (but: sound symbolism when the name sounds alike to sounds object produces)
- words and sentences can be analyzed semantically - 5pragmatic knowledge?- ability to use language in a contextually appropriate fashion (how context contributes to meaning?)
- deixis: words and phrases, such as “me” or “here”, that cannot be fully understood without additional contextual information (where is here?)
- speech acts: an utterance that has performative function in language and communication, words can be used not only to present information but also to carry out action
- implicature: what is suggested in an utterance, even though neither expressed nor strictly implied - 6language processing?- producing/comprehending (lexical access, synaptic parsing..)
- incrementality: a key word in language processing, integrates incoming different cues rapidly and updates, we make predictions up from some point because we prefer some things (subject first, main clause..), influenced by frequency, plausibility, thematic fit
- modularity: not very modular, high degree of interactivity, embodied language processing (motor and speech areas share representational subtracts, ergo not strictly divided)
- complexity: there can be a lot of embeddings which makes it difficult to parse
- ambiguity: lexical (word category/meaning), structural (word order, attachment ambiguity -- who is with the gun?, thematic role), referential, local/global, disambiguating information can only resolve local, multiple ambiguity - 7levelt model of speech production?- Processing is divided into several stages and is serial (one stage must be complete before the next can start)
- activation of lexical concept (choose the idea), select a lemma (words close to meaning of what you want to express), then retrieve it's morphological make-up (how does it behave in a sentence?), phonological encoding (metric structure -- number of syllables, phonetic encoding (producing a gestural score to coordinate the gestures), articulating
- The speaker self-monitors – listening, at a relatively low level of attention, to their own speech. When the slip is heard, the information is fed back into the conceptualiser which can then plan a correction if it is felt necessary
https://en.wikiversity.org/wiki/Psycholinguistics/Models_of_Speech_Production#The_Bock_and_Levelt_Model
OR
Conceptualisation
deciding on the message (= meaning to express)
Formulation
turning the message into linguistic representations
Grammatical encoding (finding words and putting them together)
Phonological encoding (finding sounds and putting them together)
Articulation - 8evidence for levelt?- speech errors (partial breakdown)
- tip of the tongue (only part of the process is disrupted -- likely a temporarily disruption of phonological activation, speakers can often identify the syntactical, and even morphological properties of the intended word, but still not be able to encode the phonology)
- naming influenced by production
frequency
- semi-independence of semantic and phonological processes - 9dell model?- parallel! speech is produced by a number of connected nodes representing distinct units of speech (i.e. phonemes, morphemes, syllables, concepts, etc.) that interact with one another in any direction, from the concept level (Semantic level), to the word level (Lexical selection level) and finally to the sound level (Phonological level) of representation
- word is selected, all of the nodes representing that words constituent morphemes, phonemes, semantics and syntax activated, activation spreads to the adjacent nodes until one node that is most highly activated is selected for the output string. nodes within the same level or representation competing for activation, the wrong node is often activated. with “Put your boots on!”, one would certainly think that the speaker meant “Put your boots on your feet!”. As a result, may have activated the concept nodes that are semantically related to the target word boots, such as the node representing feet. --> put your feet on - 10speech perception?- coarticulation (sounds are influenced by preceeding and following ones), allows us to speak smoothly, no one-to-one mapping of sounds to phonemes
- perceptual constancy (d will be perceived as such no matter what is following)
- can spot the difference when presented in isolation - 11motor theory of speech perception?- relating sound to mouth movement and mouth movements to phonemes
- need to be able to produce the sounds to comprehend them, brain areas involved in production are also activated in perception
- mirror neurons? but infants can distinguish sounds without producing them
- mcgurc: visual gestures affect perception --> multi-modality - 12duplex perception?- Listeners hear a syllable in the ear that gets the base (B), but it is not ambiguous. Its identification is determined by which of the nine F3 transitions are presented to the other ear (A).
Listeners also hear a non-speech "chirp" in the ear that gets the isolated transition (A)
- same stimulus is simultaneously part of two distinct types of percepts --> percepts are produced by separate mechanisms, or modules, that are both sensitive to the same range of stimuli
- also for non-speech (against modularity) - 13which brain area is important for language?left perysylvian cortex (Broca, Wernicke, auditory, articulatory motor & somatosensory)
- 14aphasia?- focal lesion, language deficit, appears after language acquisition
- multimodal (reading, writing, repetition, understanding, producing, articulation, syntax..)
- correspondence not always simple and easy
- - 15Voxel-based lesion-symptom mapping?Lesions within the insula and deep parietal
white matter had the most impact on fluency, whereas injury to the
middle temporal gyrus produced the largest effect on measures of auditory
comprehension --> perysylvian connectivity changes (stronger connections, shorter path length through jumping connections between articulatory and auditory areas) - 16what is special about the wiring of the human cortex?Dorsal fronto-temporal connection
(fasciculus arcuatus) strong and
left-lateralised in humans only - 17word learning?word is heard: activation in Superior-temporal areas, word is spoken: Inferior-frontal activation, action-perception
circuits for speech sounds and words --> action perception circuit - 18so why are perysylvian areas so important for learning?- babbling + speech production: correlation learning --> strong and short connections between frontal and temporal areas, especially at the left
- 19segregation or interaction?pulvemüller thinks: widespread near-simultaneous processing, /interaction between action and perception
systems - 20neuropsychological / psychophysiology (language physiology)?1. something happens in the brain, how does the person behave? (tms, tdsc)
2. something changes in behavior, what happens in the brain? (fmri, pet, eeg. meg) - 21what can we measure with eye-tracking?- length of fixation, how fast was the eye movement to the object, number of fixations... --> language comprehension
- 22active/passive tasks?- click on an object.. overt response
- look and listen --> predictive and interference processes (explicit/implicit) - 23tanenbaum?Visual context influenced spoken word recognition, Visual context influenced spoken word recognition -> longer until first fixation on the right word when a competitor, before that gaze between both of them
- 24CIA?asserts that incremental
linguistic interpretation actively directs attention in the visual environment, thereby increasing the salience of
attended scene information for comprehension. - 25how does eye-tracking work?- at first a lot of weird techniques
- then corneal reflections: to take a photo of the reflection on the fovea which an external light source produces
- purkinje images: they are reflections on the boundaries of lens and cornea when infrared light shines into the eye, two relative positions: image from the outer cornea and reflection of light off the retina (bright eye) -> direction of gaze calculate its movement relative to these points - 26eye movements?- nystagmus
- saccades: high velocity, no information acquirement
- fixation: eye rests and takes in information (200-500 ms)
- fovea - hughest acuity, 2 degress around center, parafovea: 5 degrees
- - 27reading?- 3-4 characters left of fixation, 7-8 to the right, word skipping more likely for short and function words
- fixation for the first or second time? re-reading context on the left?
- first pass ( time spent in a region before
moving on or looking back), regression path (time from first entering a
region until moving the eyes beyond that region, includes
regression time), total time - 28phonemes?- smallest sound particles that distinguish between meaningful words (know bilabial, nasal, alveolar...)
- 29phonemes and the brain?lateral, aneterior and posterior superior temporal lobe are active when phonemes vs random sound, phonemes vs speech sounds of other languages, specific phoneme categories
- 30mmn and phonemes?reflects pre-attentive memory traces for speech sounds of the native language, develops for another language as you learn it
- 31action-perception circuit?speech-induced articulatory and acoustic
phonological activation create action-perception circuits (when I talk, I hear the sound I make so that neurons producing the sound become connected to the neurons representing the acoustic form of the sound), inferior frontal and motor areas activate when speech is heard, also back and forth communication, activation in the motor cortex specific for how the sound you heard was produced (for movement, articulation and speech perception) - 32role of sesorimotor areas?- motor areas play a causal role, same motor regions activated by hearing the sound, pronouncing it with and without sound
- reflects different phonemes / distinctive features
-> action-perception - 33naming?Simultaneous brain manifestation of word
frequency and phonology (place of
articulation) at 200 ms post picture onset --> simultaneous ignition hypothesis - 34what's in the lexicon?form, meaning, use
- 35how do we recognise words?- very fast although near 65000 words some of which compete, selection happens very early
- errors are rare
- retrieval and not a compositional task
- - 36embedding problem?words resemble other words, other words may be embedded within it, words may be embedded in word sequences
- 37Characteristics of lexical processing?- incrementality (recognising by unfolding)
- several competing activated in parallel
- others are inhibited as the winner becomes more clear
- acoustic-phonetic information can modulate lexical processing - 38cohort model?Retrieving lexical
information began
prior to word offset (fixations between competitors) - 39uniqueness point?when no other word in the
language matches the string of phonemes in the mental lexicon, we can recognise a word, sometimes you still need the word offset, 100-200 ms after this point eyes start to move to the target - 40competition and embedding?word recognition depends on how many competitors, also parallel competition of whatever rhymes, not only words starting with a sounds, but also the ones that contain it
- 41phonetic information?listeners make immediate use of anticipatory coarticulation in the determiner to predict the initial sound(s) of the upcoming word --> faster than when it's neutral (schwa in the determiner undergoes anticipatory vowel to vowel co-articulation)
- 42other?frequent words are recognised faster than infrequent words, activation of semantically related words
- 43Models of word recognition?agreement: a lot of words activated in parallel, activation depends on the match between competitors and speech signal, activation of competitors during recognition
differences: how does information flow? what is the nature of the different representations? abstractness - 44cohort?verbal, mapping sound information to the lexicon, strong emphasis on word onsets, three steps: activation of representation based on sounds, selection, integration
cannot account for frequency effects and recovery from early errors --> if it sounds a bit different from the word stored in the lexicon we are still able to contextually recognise it and include in the cohort)
cohort ii: fully bottom-up, words that have a minimal mismatch still can enter the cohort
frequency: higher resting activation value
activation: depends on the goodness of fit between representation and speech - 45trace?parallel processing of sub-systems (take all of the various sources of information found in speech and integrate them to identify single words), bidirectional (all layers can have influence), connections within are inhibitory and between fasciliating, Higher-frequency words have higher baseline levels of activation, reach a critical activation threshold for recognition more quickly than low-frequency words
Word activation does not decrease for mismatching input, lexical candidates that are sufficiently similar will be also activated even without same onset (but people look more at onset competitors) - 46fine tracker?effects of fine phonetic detail, a computational model, acoustic signal is an input, 1 level: an artifical neural network transforms the waves into feature vectors, words represented as articulatory feature vectors, probabilistic word search
- 47word categories in the brain?content and function words:
differential laterality of ERP brain responses at 160ms, In order to elicit left lateralized cortical activation a rapidly changing spectrally rich sound with phoneme status needs to be placed in the context of frequently encountered larger language elements, such as words.
morphemic processing seems to be supported by distinct neural networks, thereby providing evidence for a specific neuronal basis for morphology as part of the cerebral language machinery, different
with regard to their cortical substrates - 48more of it?semantic meaning is widely distributed, learning the meaning of words related to
internal states, e.g. pain; dependant on meaning = different areas in the brain, semantic symbol representations through unique corrrelations, semantic hubs = correlation learning, Semantic
somatotopy, semantic category
differences - 49n400?most over parietal, semantic mismatch, doesn't matter for syntactic violations, a bit earlier and stronger in the sentence context, but latency and distribution not differentiable
clear effect of semantic congruity across modalities (also line drawings and rebuses...) less when repeated, larger negativity in first words of a sentence --> more unclear expectations in the beginning
- N400 indexes both storage and combination (if the word doesn’t fit the context = don’t combine well AND if we use pseudo words = not stored) - 50graded semantic processing?if words share initial phonemes, n400 starts later, also more for beetwen categories violation (tulips/palms)
- 51coercion effects?noun less predictable (begin the book vs begin reading the book)
- 52verb-action mismatch?larger negative amplitude for mismatching trials, centro-parietal maxium,
- 53take home on semantics?visual and context info rapidly integrated, similarity between linguistic and visual contexts, preferential reliance on reference (and not knowledge!), spatial distance play a role, eye gaze modulated by motion verbs
- 54different views on storage/combination?- classical models: words are stored as a whole, as well as rules; rules combine them into phrases/sentences
- cognitive/construction grammar: all language is stored as constructions (plural = N+s etc) -> constructicon - 55criteria for deciding whether stored/combined?- how frequent is it used? (if not so much -> rather combined)
- semnatic opaqueness (the more transparent, the more likely stored)
- maybe brain measures can contribute?
- - 56linguistic access in comprehension: serial or parallel?- different models:
- ELAN: syntax first/serial model, some studies found ELAN and NO N400 (serial, doesnt go past this stage), other models found both (parallel)
- P600: syntax-related (amount of time and effort in general it takes to build up coherent structure in a sentence? or reanalysis?)
- serial: first syntax, then semantic
- parallel: very early brain responses to distinguish on all dimensions, N400 is a second-order response, 100-200 ms (different features of meaning firstly assessed early: word semantics, difference between literal (compositional) and idiomatic (whole world form) meanings) in semantic hubs (pFC and anterior temporal lobe) - 57whole form storage?- action perception circuit: correlation between production and perception of symbols/sounds in motor and sensory areas
- for speech sounds, words, frequent words sequences
- storage: strongly correlated elements in a network?
- words = networks in the perysylvian area? strong brain response for meaningful words and reduced for pseudo words (the whole network cannot be activated); - 58MMN for perysilvian traces for words?- weaker MMN for pseudowords in temporal
- 59mechanisms for combining words into chunks?- correlational firing leads to the storage of generalising rules and individually stored constructions (word-to-word combinations)
- articulatory + auditory neurons bound through AF
- generalised schemes about whatever goes together as wired together through correlations
- a matrix with words that go together (connections are strenghtened), also co-activation of other strong correlations —> strongly connected conglomerate of cells = rules (as this is strongly correlated this must be the rule that it always goes like this)
- computational: network learned the co-occurence: generalising — if this word has been occurring in the context of the word from matrix, then its also likely to combine with the items which have been combined with this original context word (other animate words have been combined with words from this part of the matrix, so a new animate word would also be, combinatorial neural assembly (basics for rules and generalisations) develops
- sentence structure: V-N, after a noun congruent verbs are primed and thus less activation then incongruent - 60neural model of grammar?- enhanced MMN to grammatically incorrect strings compared to grammatical strings;
- lexical storage was the other way around (good ones produced larger MMN), but larger MMN for syntactically incorrect strings —> different signatures of whole-form and combinatorial processes
- whole-word enhancement of MMN, combinatorial depression of MMN
- predictions and prediction errors are of two kinds: whole form prediction and combinatorial prediction, lexicon and syntax needed - 61Are morphologically complex words stored as whole forms?- MMN: lexical pattern (MMN larger to the good form, storage)
- 62are particle verbs whole-formed stored?- MMN larger to meaningful verbs —> stored, whole form storage pattern
- 63summary?- symbol meaning is accessed immediately
- whole symbol access and combination are parallel distinct processes
- simultaneous constructional (lexical) and
combinatorial semantic processing of idioms - 64fodor's modularity?- modules are domain-specific, informationally encapsulated, fast, autonomous, innately specified, neurally specific, not assembled
- three levels: trnasducers convert stimuli into neural signal, input systems are modular, interpret transducer information and responsible for basic cognitive activities, central system not modular, complex cognitive functions - 65garden path theory?- model of sentence processing
- parsing decisions: minimal attachment: syntactically simplest analysis (defined in terms of nodes) is preferred
- high-attached analysis (attached to a node which is higher in the syntactic tree)
- late closure: the principle that new words tend to be associated with the phrase or clause currently being processed rather than with structures farther back in the sentence - 66frazier: how parsing goes?- serial structure building
- reanalyzing based on syntactic or semantic (low plausibility) conflict
- this processing is innate
- processing is modular as only syntactic information is used for structure building - 67against modularity?- evidence of very early interaction of frequency, animacy, plausability...
- computational frameworks: competitive activation, connectionist systems
- rapid online integration of semantic, world knowledge etc information --> parallel interactive account of comrehension - 68linking hypotheses with reading?- amount of time spent in a region reflects ease/difficulty of processing
- first pass time: early processes
- second/total pass: late processes
- regression: unclear, integration of left information - 69constraint-based processing?- sentence processing involves immediate incorporation of all available information in creation of the final output
- semantic meaning of the verb influences the assignment of thematic roles very rapidly --> biased by an animate object which seems to be an agent, an inanimate object resolves the disambiguity quicker - 70visual scenes?- visual referential context helps resolve local ambiguity (apple on the towel)
- world knowledge allows for incremental thematic role assignment (when you hear der hase frisst... you instantly look at the cabbage, not at the fox), resolvement of syntactic ambiguity
- depicted events cause -> rapid incremental structural disambiguation + thematic role assignment
- visual context influences syntactic structuring - 71can affordances help with syntactic ambiguity from the very beginning?- yes
Information
- Description
- Flashcards on Language and the Brain, created by Lisa Ka on 07/06/2016.
- Created by
- Lisa Ka
- Views
- 1
- Rating