Cognitive Psychology: Language, Memory, and Speech Perception, Study notes of Psycholinguistics

Psycholinguistics notes for linguistic majors and others

Typology: Study notes

2018/2019

Available from 02/06/2022

miska-vlcejova
miska-vlcejova 🇸🇰

12 documents

1 / 9

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Psycholinguistics
-an interdisciplinary field of study (psychology and linguistics)
-Language in the mind = Studies mental representations and processes in language use
How do people learn a language (or two)?
How do people use language to understand each other?
How language is represented and processed in the mind? (and the brain – neurolinguistics)
-Two underlying questions:
1 What knowledge of language is needed for us to use language? = Linguistics
2 What cognitive processes are involved in language use? = Cognitive psychology
-Language:
1Phonology - Language sound systems
2Semantics - Meanings of words and sentences.
3Syntax - How words are arranged to form a sentence.
4Pragmatics - Social rules in language use.
-Cognitive processes:
a)perception
b)memory
c)thinking
-Research methods in Psycholinguistics:
a)Observation
b)Experiments
c)Neuroimaging studies
The Components of Psycholinguistics
-Developmental psycholinguistics: How language is acquired during development? (1st, 2nd, in childhood or later etc.)
-Experimental psycholinguistics: How people comprehend and produce language? (both in speech and in writing)
-Applied psycholinguistics: How psycholinguistics can be applied to other areas of life?
-used in pedagogy, learning, language development, speech therapy, criminology, computational linguistics (model language use
in PC)
Cognitive Psychology
-It is the study of perception, attention, memory, language, and thinking in humans...how we know about the world.
language itself is its object
It studies processes that play a crucial role in lg. learning and lg. use.
-Another perspective: the scientific study of the human mind and information processing
Information processing
-Language processing = information processing
-Information processing model = computer analogy
Sensory Memory
-not conscious!
-Information from the five senses
-Sensory register
-Large capacity
-Short duration – lot of inf lost
-Role of attention and perception
-Person #1: "What time is it?" Person #2: "What did you say? Oh, 2:30."
Attention
-selectively concentrating on one aspect of the environment while ignoring other things.
-Or: the allocation of processing resources.
-Function: selects information and patterns that make their way to STM / WM
-Focused Auditory Attention
many conversations in the background, but only listen to one.
distinguishing: picking up on a number of features: Speaker’s voice Speaker’s location Content of speech
Divided attention:
oEasy for some things: Driving and talking
oHard for other things: Patting head and rubbing stomach
Factors affecting dual task performance
-Task similarity:
-Easy to do things which are dissimilar (drive & talk).
-doing different things with items in the same modality = difficult
How can you multi-task?
-some processing becomes automatic (experts!)
-Automatic processes do not tax attention:
oFast
odo not reduce capacity for performing other tasks
ounavailable to consciousness
ounavoidable
-Language processing!
What is the difference between sensing and perceiving?
-Perception = organization, identification, and interpretation of sensory information in order to represent and understand what we
have seen, heard etc. – pattern recognition
-Both visual and auditory
1
pf3
pf4
pf5
pf8
pf9

Partial preview of the text

Download Cognitive Psychology: Language, Memory, and Speech Perception and more Study notes Psycholinguistics in PDF only on Docsity!

Psycholinguistics

  • an interdisciplinary field of study (psychology and linguistics)
  • Language in the mind = Studies mental representations and processes in language use
    • How do people learn a language (or two)?
    • How do people use language to understand each other?
    • How language is represented and processed in the mind? (and the brain – neurolinguistics)
  • Two underlying questions: 1 What knowledge of language is needed for us to use language? = Linguistics 2 What cognitive processes are involved in language use? = Cognitive psychology - Language: 1 Phonology - Language sound systems 2 Semantics - Meanings of words and sentences. 3 Syntax - How words are arranged to form a sentence. 4 Pragmatics - Social rules in language use. - Cognitive processes: a)perception b)memory c)thinking - Research methods in Psycholinguistics: a)Observation b)Experiments c)Neuroimaging studies The Components of Psycholinguistics
  • Developmental psycholinguistics : How language is acquired during development? (1st, 2nd, in childhood or later etc.)
  • Experimental psycholinguistics : How people comprehend and produce language? (both in speech and in writing)
  • Applied psycholinguistics : How psycholinguistics can be applied to other areas of life?
  • used in pedagogy, learning, language development, speech therapy, criminology, computational linguistics (model language use in PC) Cognitive Psychology
  • It is the study of perception, attention, memory, language, and thinking in humans...how we know about the world.  language itself is its object  It studies processes that play a crucial role in lg. learning and lg. use.
  • Another perspective : the scientific study of the human mind and information processing Information processing
  • Language processing = information processing
  • Information processing model = computer analogy Sensory Memory
  • not conscious!
  • Information from the five senses
  • Sensory register
  • Large capacity
  • Short duration – lot of inf lost
  • Role of attention and perception
  • Person #1: "What time is it?" Person #2: "What did you say? Oh, 2:30." Attention
  • selectively concentrating on one aspect of the environment while ignoring other things.
  • Or: the allocation of processing resources.
  • Function: selects information and patterns that make their way to STM / WM
  • Focused Auditory Attention  many conversations in the background, but only listen to one.  distinguishing: picking up on a number of features: Speaker’s voice Speaker’s location Content of speech Divided attention: o Easy for some things: Driving and talking o Hard for other things: Patting head and rubbing stomach Factors affecting dual task performance
  • Task similarity:
  • Easy to do things which are dissimilar (drive & talk).
  • doing different things with items in the same modality = difficult How can you multi-task?
  • some processing becomes automatic (experts!)
  • Automatic processes do not tax attention: o Fast o do not reduce capacity for performing other tasks o unavailable to consciousness o unavoidable
  • Language processing! What is the difference between sensing and perceiving?
  • Perception = organization, identification, and interpretation of sensory information in order to represent and understand what we have seen, heard etc. – pattern recognition
  • Both visual and auditory
  • Why are they important in language processing? Schema theory
  • We actively process information in our world. – use our schemas to recognize patterns in the world
  • If information is missing, the brain fills in the blanks.
  • This can result in mistakes – distortions
  • Auditory: think of lyrics (pop music) What is STM? (Short-term memory)
  • The part of memory where you store information at the present moment / for a very short time (some seconds!) for conscious processing
  • Similar to RAM in computers
  • Common misconception: STM = what you remember from the past 2 days
  • it contains information one is currently thinking about or has recently thought about
  • within 30–60 seconds, some say: only 15 sec
  • Capacity: 7 ± 2 items (Miller’s law) Some say: more realistic (4 ± 1) chunking
  • Function : - Maintenance of current information, often by rehearsal (e.g. saying a phone number over and over again.) - Problems with the classic STM-model: o Unitary system : research showed that some people – problems with auditory STM but not with visual (or vice versa) o Does not explain processing (e.g. mental arithmetics) The Working Memory model - Central Executive
    • Does work in the WM: o allocates attention o selects buffers for information o transforms and manipulates inf a) Phonological Loop - Holds auditory info
      1. Articulatory Loop - Rehearsal: inner speech – AS or LTM
      2. Acoustic Stor e - Simple storage of acoustic info (decay after 2 sec) b) Episodic Buffer - Holds and combines chunks of information from PhL, VSS and LTM c) Visuo-Spatial Scratchpad - Holds visual and spatial inf. - Testing working memory
    • Forward and backward digit span
    • Forward and backward syllables/words/non-words
    • Remembering nonsense words that get longer
    • Serial order is important, too!
    • 1-back, 2-back, n-back tasks
    • Quite a few other techniques exist! - The significance of WM
    • Correlates with o Reading comprehension o (mathematical) problem solving o IQ o Success of foreign lg learning But how do we transfer information from WM to LTM?
  • Atkinson and Shiffrin proposed Rehearsal (repeating information)
  • Sometimes true, but not always supported by experiments
  • Too simple
  • Also: people with average WM and outstanding LTM….. ???
  • Why is it then that people with large WM often have an LTM advantage? o WHAT they do with the information counts Levels of Processing Theory
  • The deeper we understand information the more we retain.
  • Different ways of processing will aid recall
  • People with „good” LTM – engage in deep processing o Process the same information in several ways o Easier for people with large WM o You can do it with average WM too o Family background important
  • Important in language learning (vocabulary) Long Term Memory 1 Explicit (declarative) a)Semantic memory - facts, general knowledge (network) b)Episodic memory - personal experience 2 Implicit (procedural) a)Motor skills b)Cognitive skills c)Conditioned responses, habits Semantic memory
  • Semantic network: semantically related concepts are linked
  • „relatedness = co-occurance”
  • propositional networks: o Representing factual statements in the mind o Events, relations can be represented

2 Positional processing Grammatical encoding – positional processing

  • Organize lemmas into an ordered string 1 Constituent assembly creates a sentence frame for the message: (determiner) N 1 V[past] (determiner) N 2 2 Lemmas for the content words are accessed and inserted into the frames 3 Function words and grammatical endings specified 4 Lexemes accessed à phonetic plan generated Evidence for the distinction between functional and positional processing
  • Comes from speech errors:
  • You must be too tight for them. X
  • (They must be too tight for you.)
  • Wrong functional roles have been assigned to the pronouns. (functional processing error)
  • It is not a simple ordering problem!!! = Grammatical marking is appropriate
  • A simple ordering error would be: o You must be too tight for they. Finding words
  • Transforming thoughts into words
  • Two levels: 1 retrieval of the abstract word form (lemma) 2 retrieving the form of the actual word (spoken – phonological form; written – visual form)
  • Separation supported by brain imaging studies What do pauses tell us about lexical retrieval?
  • Indicate retrieval problems
  • Some words are easier to retrieve: frequency
  • Another factor: predictability (depends on context)
  • The two are not the same! E.g. soup – broth: Which one is more frequent?
  • But: „Too many cooks spoil the ……”
  • Pauses are more likely and longer if the word is rare and/or unpredictable Measuring the likelihood of content words in an utterance
  • Shannon guessing game: try to guess the next word in the sentence. Number of guesses counted
  • „Look back a year, and remember how disquieting European politics... „
  • Then go on to the next word etc.
  • The lower the ratio of correct guesses to incorrect guesses for a word, the lower the likelihood of that word in that context. Speech errors – another source of data
  • Shed light on the mechanisms of word selection
  • Cause (being tired, anxious etc.) not interesting for psycholinguist
  • Mis-selection is relevant for us now (substitution and blends) Substitutions and blends
  • Involve words that are semantically related
  • Substitutions: - one word replaces another likely to involve antonyms
  • Blends - two words are merged
  • Synonyms or near-synonyms Levels of processing different for the two
  • Blends : concept level problem o two related concepts activate two related lemmas o Competition in speakers mind – competition not resolved
  • Substitution: lemma level problem o Remember semantic memory? o Related words stored next to each other – for some reason, another (neighbouring) lemma gets activated o Possible reasons: a) it is more frequent b) it is easier to imagine c) target word temporarily unavailable (inhibition) Malapropism
  • The word produced is similar phonologically to the target word but has a different meaning. - RARE
  • It is only a speech error if the word with the correct meaning was intended!!!
  • If these two vectors are equivocal, then…
  • If these two vectors are equivalent, then…
  • Target word gets activated – its component sounds too à component sounds activate the wrong words
  • Reasons behind: frequency, context (words fits in context, already activated to some extent) Morphological and phonological encoding
  • Interestingly, words must be built up: o Morphologically (put together meaningful parts) o Phonologically (put together sounds)
  • How do we know? o Tip-of-the-tongue (TOT) o Other speech errors Tip-of-the-tongue
  • Lemma activated, but word form cannot be constructed (for some reason)
  • Give subjects definitions of uncommon words – TOT often induced
  • Subjects could often tell: o The beginning/end of the word o Number of syllables o Which syllable is stressed o Rhyming words o Grammatical gender Inferred assumption
  • There is word-sketch we should fill when we activate words
  • TOT: some essential parts unavailable (phonemes and their exact order) What about words consisting of several morphemes?
  • Are affixed/suffixed words constructed or stored as a whole?
  • She already trunked two packs. X
  • She already packed two trunks.
  • Sentence structure indicated where inflectional morphemes should be inserted.
  • If words mixed up, then suffix attached to wrong word.
  • à free and bound morphemes stored separately, then put together Difference between derivational and inflectional morphemes
  • Inflectional morphemes: verbs, nouns, adjectives with a marker
  • Derivation: affixes and suffixes to form new words
  • Any guesses?
  • Inflections: stored and activated separately
  • Derivations: more controversial – many of them stored as a whole (particularly complex ones, and where meaning is not transparently related to stem – e.g. accountant)
  • But: in an experimental situation subjects can form the opposites of non-words!
  • Also: there are speech errors when we use an incorrect prefix: inclear* (unclear) How we understand spoken language? „Mirror of production” 1 Reconstructing phonological representation 2 Retrieving lexical items 3 Create syntactic representation (parsing) General difficulties
  • Variability: the „same” phoneme is actually different, depending on who produces it and when + the phonetic context
  • Extreme problem for computers!
  • How can we recognize the phoneme [a] always as an [a]???
  • Not one template but several exemplars are stored in LTM – speaker characteristics (age, sex, socioling. Background) + phonemic environment
  • Also: both bottom-up and top down processing used by humans! Basic issues in speech perception
    • Human hearing system is best in the frequency range of speech sounds (600-4000HZ) – we are tuned to speech
    • The mind treats language and non-language signals separately (immediately differentiates between the two)
    • Phoneme restoration effect
    • Cough is treated (and interpreted) as a separate stream of sounds Phoneme restoration effect
  • Stronger with real words than with nonsense words
  • If word level information ambiguous, the word matching the sentence context is restored
  • E.g. It was found that the [#i:l] was on the table. (heel, meal or deal?)
  • It is also a proof of top down processing! Multiple cues to speech sounds
  • Several auditory signals help us identify a phoneme:  Characteristics of the phoneme (voicing, silence during lip closure - plosives like [p] etc.)  Properties of the burst noise (in plosives) etc.  Changes in the characteristics of the preceding vowel  Changes in the length of the preceding vowel
  • Cues must be integrated
  • Sometimes visual cues – McGurk effect
  • Haptic (touch) cues – puff-of-air effect Coarticulation
  • Neighbouring sounds influence each other: articulation of one sound is dependent upon what sound preceds/follows it
  • What do you hear? Tapes or capes?
  • [t] and [k] manipulated artificially to get a list of in-between sounds
  • Subjects presented with Christmas [#apes] and foolish [#apes]
  • When in-between sounds presented, the phoneme identified by subjects depended on what the preceding word (last sound matters) was! (same sound!!!) Categorical perception
  • If sounds on a continuum presented, we tend not to hear e.g. 5 different sounds, but 2 (we put them into two categories)
  • Identification and discrimination tasks
  • Consonants can be placed on a continuum: [ba] and [da] generated by computer So how do we make sense of all this? (Summary)
  • Question : When do we analyze the syntactic structure of a sentence? When the speaker has finished the sentence?
  • NO – it would put an incredible load on WM (some sentences 30-40 words long)
  • Incremental processing: We start processing as soon as information arrives o Build hypotheses à modify if necessary Ambiguous sentences
  • The man hit the dog with the leash.
  • Retired priest may marry Springsteen (authentic headline)
  • Panda maiting fails; veterinarian takes over (authentic headline)
  • Kids make nutritious snacks (authentic headline) Several approaches to sentence processing 1 Serial (modular approaches ): a)Build one model based on syntactic information only b)Try to modify it as long as possible 2 Interactive analysis : use both semantics and syntax to build the „best” structure 3 Parallel analysis : build more than one alternative structure at the same time. Parsing principles
  • Seem to be universal, but applied in a different way in different languages
  • Agglutinating languages (Hung): primary analysis – affixes and inflections, secondary – word order
  • Analysing languages (Eng): primary – word order Summary 1 Reconstructing phonological representation:
    • Overcoming difficulties (of variability)
    • Several characteristics of human speech recognition (phoneme restoration, McGurk etc.) 2 Retrieving lexical items (Cohort model) 3 Create syntactic representation (parsing)
    • Serial, interactive and parallel models Reading and literacy, Psycholinguistic processes
  • How do our eyes move when we read?
  • Is it a smooth, continuous movement? No
  • „jump” (= saccade) from one fixation point to the next.
  • Louis Émile Javal (19th) c.
  • Saccade: jump – nothing is seen – 20-40ms o 7-9 character jump (up to 20)
  • Fixation: 100-500ms (avg: 200-250), 4-5 letters + peripheral vision Eye movements
  • The eye behavior is constrained mainly by the cognitive processes and visual influence.
  • The eye tracking technique provides a possibility to study the reading processes in an ecologically natural situation. Letter recognition
  • Template matching theory: there is a template in our mind (‘A’) and we match the incoming stimuli to that template. (one-to-one match) Pattern recognition
  • Prototype theory : central feautures must match but there can be „peripheral” differences. – flexibility
  • Bottom-up processing
  • What you have seen in the words: top-down processing
  • Let’s move on to words. What does this demonstrate?
  • We (experienced readers) do not read letter by letter, building up words from individual letters Models of Word identification 1. 1 Dual Route Cascaded (DRC) model 1 Two ways a word’s pronuniciation can be generated: a) grapheme-to-phoneme conversion (with the help of rules) b) Direct mapping of spelling to pronunciation (sight-reading, whole word recognition) 2 Both the orthographic and phonological forms of words are represented holistically in the lexicon; the assembled and direct routes operate in parallel → pronunciation of a word jointly determined by both routes Models of word identification 2.
  • The connectionist model – will not cover in detail
  • Word features processed at the same time in the mind Sentence processing
  • When Fred eats food gets thrown.
  • The girl told the story cried.
  • I convinced her children are noisy.
  • The horse raced past the barn fell. Etc.
  • „Garden-path sentences” = grammatically correct sentences that start in such a way that a reader's most likely interpretation is an incorrect one What happens when you are led down the garden-path?
  • While Anna dressed the baby spit up on the bed.
  • You start along a path = construct a syntactic structure
  • Spit = „stumble” – regressive eye-movements (does not fit the structure you built)
  • Reanalysis begins A modified version of the same sentence
  • While Anna dressed the baby that was small and cute spit up on the bed.
  • Reanalysis is more difficult if the head of the misanalysed phrase is distant from the error signal!
  • The length of the sentence is not important.
  • Many people do not understand these sentences! Models of syntactic parsing 1.
  1. Garden Path (modular/serial) Models : give priority to the grammatical structure of the Sentence.
    • A single parse is constructed – then revised if necessary
    • Two basic parsing principles: o Minimal attachment – prefer interpretation with the simplest structure (next slide) o Late closure („Don’t start a new clause!” [unless it’s absolutely needed) Minimal attachment
  • What is simplest?
  • Fewest branchings in the tree – count the number of nodes! Models of syntactic parsing 2.
  1. Constrained based (interactive/parallel) models : gramm. Structure is just one factor influencing S interpretation (others are plausability and appropriateness - semantics) „The cop arrested by the detective was guilty of taking bribes.”„The crook arrested
    • Not clear which model is more appropriate
    • Can be the combination of the two Models of Discourse Processing
  • Describe processes that are necessary to connect the meaning of individual sentences
  • Build upon representations provided by the word-identification and the syntactic parsing processes
  • Kintsch-Van Dijk: construction – integration model The construction-integration model 1 Construction: the literal interpretation of the text created (text-base) additional, related info retrieved from LTM All these integrated into a network of propositions - done on 1 phrase or S at a time – no. of propositions dependent on WM capacity 2 Integration: associations among important propositions strengthened, among less important ones – weakened → textual inconsistencies minimized, less important information eliminated etc. What you need
  • What we have not stressed yet: o Prior knowledge o Reading goals o Discursive knowledge (frameworks and scripts, discourse or story structures) Inferences (while reading)
  • Forward/predictive: anticipate events or outcomes
  • Backward /bridging The role of reading skills and attention
  • Attention – limited capacity
  • If attention is required to identify individual letters à less attention available to recognizing words
  • If attention is required to recognize words à less attention available to building up story context
  • Key – automaticity (remember the Stroop effect?)
  • Less experienced readers: rely even more on context than experienced readers (research!) but they „forget” it, because they have to focus on word decoding! Summary
  • Eye-movements Letter recognition
  • Word recognition Sentence parsing
  • Comprehension Language Development Theories Introduction
  • Children begin developing language at birth as they interact with their caregivers.
  • Children from culturally and linguistically diverse backgrounds are exposed to diverse experiences; this diversity creates the concepts to which children attach symbols, or words. The role of the environment in language development
  • Development of language rests upon several major variables that interact with one another
    1. The child’s cultural and linguistic environment is a big influence upon language learning
    2. Each child has unique characteristics that she brings to the language learning situation Behaviorist Theory : a)Imitation theory
    • Explains acquisition of verbal behaviour
    • Imitation is key
    • Cannot explain children’s errors
    • Cannot explain how we can generate new sentences b)Reinforcement theory
    • Verbal behaviours are learned under appropriate conditions of stimulation, response, and reinforcement (praise or ignore)
    • But: Parents seldom correct children
    • They often check the truth of the statement and not its form.
    • When they do correct, it is no use. Theories of Language Acquisition