PSYC 211 Fall 2024 Language PDF

Language PSYC 211 Fall 2024  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Learnin...

Language PSYC 211 Fall 2024  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Learning outcomes Provide a general definition of language Describe and give examples of important structural components of language Identify what makes language hierarchical Define and distinguish syntax from semantics  What is language? More than just speech! Complex, hierarchical system of communication Symbolic and arbitrary (Woodward & Hoyne, 1999) Words do not map to objects in a literal sense Sign language = full language Cat Chat Kissa  Basic structural components of language Phonemes = basic unit of sound in a language /R/ai/n/b/ow/s Syllables = small combinations of phonemes Rain-bows Morphemes = smallest meaningful unit of speech Rainbow-s Words = combination of phonemes, morphemes, Rainbows and syllables Phrases, Sentences = combinations of words Rainbows are pretty Five Levels of Analysis  Basic rules govern how we combine words Syntax = rules for how words are combined in language Structure of phrases, governed by grammar Ex: The black cat was sitting in the window. Ex: The cat black was sitting the window. Semantics = meaning of language Ex: The black cat was friendly. Ex: The black backpack was friendly.  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Learning outcomes Define and compare the two major views on language acquisition Explain different sources of evidence for the Nativist view, including examples Explain a recent challenge to the Nativist view, including an example  Major views on language acquisition Language acquisition = language learning Begins in infancy Big Q: What is required [cognitively] to learn language? 2 classical answers: 1) Only learning is required (behaviourism) 2) Some innate knowledge about language is required (nativist)  View #1: Behaviourist views on language acquisition B.F. Skinner Language acquisition is the result of trial & error learning Reinforcement and modelling No need for innate or ‘internal’ architecture That’s wrong. I amn’t “I’m not tired” I’m not Mama Mama tired tired  Views #2: Nativist views on language acquisition Noam Chomsky – I don’t think so, Skinner. Rejects behaviourist view of language acquisition Poverty of the Stimulus: Not enough exposure to specific instances of language I amn’t tired Children make inferences about how language works Universal Grammar: innate ‘architecture’ for learning language Specific to Syntax “Blueprint” of how language works Not specific to a single language  Supporting Nativisim: Pidgins and Creoles Pidgin languages: functional communication, but incomplete grammar Creole languages: come from Pidgin languages and have complete grammar Argument: Children of Pidgin language speakers develop Creole -> innate draw to build a complete grammar present in children  Supporting Nativism: Congenital Deafness and Sign Language Congenitally-deaf children not exposed to sign language spontaneously develop basic grammatical systems through gesture Goldin-Meadow & Mylander (1998) -Analysed gestures made by deaf children never exposed to sign language -Observe generative strings of gestures -Observe structural similarities between American and Chinese childrens’ gestures  Recent Challenge to Nativism: Large Language Models Computational cognitive modelling challenge to Nativism! Natural language processing: Create a machine/AI that is indistinguishable from a human Requires human-level grasp (use?) of language How do you do accomplish this? Option 1: Provide basic linguistic rules to AI (Nativist approach) Option 2: Learning!  New Challenge to Nativism: Large Language Models Input: “I love cake. It’s my favourite dessert.” Parallel processing: Examine all input at once Identify relationships between words Weak relationship It’s my favourite dessert Strong relationship I Never explicitly taught rules love of language…but can learn language! cake The debate continues!  Write down the ‘muddiest point’ for you so far What didn’t you understand? What question(s) do you have about the lecture material so far?  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Learning Outcomes Define phonological processing and the cognitive challenge it presents Describe solutions to the challenge of phonological processing Define lexical processing and explain how word frequency and context impact it Define parsing and the Garden-Path Model Explain how semantics interacts with syntax to comprehend sentences  Language is tricky… Child: Do I have to put an ant and a tiger? Dad: What? Where do you want to put an ant and a tiger? Child: No, ant and a tiger! Dad: ??? … Do you mean hand sanitiser? Child: That’s what I said!  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #1: How do we process and understand the phonemes and morphemes in a language?  Phonological Processing Reminder: Phonemes = basic unit of sound in a language Morphemes = smallest meaningful unit of speech Need to understand phonemes and morphemes to comprehend speech One major challenge: Segmenting speech signals  Phonological Processing: The Segmentation Problem Speech = continuous signal, but words have to be separated for comprehension What do you mean? How do we segment this auditory (speech) signal so that we comprehend 4 words? Gazzaniga et al. 2008  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem Statistical Learning: x1 Yeah, x1000 obviously…  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem Statistical Learning in Language: Encounter patterns in language over and over again -> encode the frequency of these patterns -> learn to detect these patterns x1000 Whatdoyoumean? What do you mean? Okay…but HOW do we do this? x1000 Whatdoyoumean? Whatd oyou mea n?  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem Transitional Probabilities to the rescue! Transitional Probability (TP): The likelihood that one sound follows another sound in speech High TP = high likelihood Low TP = low likelihood Not all sounds in speech have equal TP Sounds within a word are more constrained than sounds across words Saffran et al. (1996)  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem Whatdoyoumean English phoneme combinations: ‘do’ more common than ‘td’ within words -> ‘do’ = high TP ‘do’ and ‘td’ both common across words -> ‘td’ = low TP Sample speech over and over and over… High TP = single word Low TP = separate words Saffran et al. (1996)  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem Testing Transitional Probabilities Sronkibbakutrugswipradbakuarop… Test phase: Which word did you hear? Ex: Baku vs Baka Question: Do people segment the ‘speech’ stream Saffran et al. (1996) into words?  Phonological Processing: Statistical Learning as a Solution to the Segmentation Problem identified (max 36) Mean = 27/36 Question: Do people # Correct words segment the ‘speech’ stream Chance into words? Yes! Interpretation: Words Differentiating ‘words’ based on their TP Saffran et al. (1996)  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #1: How do we process and understand the phonemes and morphemes in a language? Problem of speech segmentation Statistical learning Transitional probabilities  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #2: How do we match words to meanings?  Lexical Processing: What is it? Lexical processing: Matching words to their meaning Fast, automatic process Challenge: Many words have multiple meanings, or words with different meanings sound the same  Lexical Processing: What cues do we use? Frequent usage: Process meaning consistent with most frequent usage of a word The baseball player threw the bat vs Frequency interacts with context The baseball player threw the bat The scared child threw the bat  Testing Lexical Processing: Lexical Decision Tasks Lexical decision tasks: See strings of letters and make rapid judgement: Word or non-word? Common word: Group Proug Less common word: Obscure Curboes Logic: Faster replies to common words = stronger mental activation/representation Tells you about lexical organisation/representation  Testing Lexical Processing: Extending Lexical Decision Tasks Lexical decision tasks: See strings of letters and make rapid judgement: Word or non-word? Use lexical decision tasks to test whether multiple word meanings are activated before we settle on a single meaning Resolving semantic ambiguity  Testing Lexical Processing: Extending Lexical Decision Tasks Rumor had it that, for years, the government building had been plagued with problems. The man was not surprised when he found several bugs in the corner of his room. Ant Rumor had it that, for years, the government building had been plagued with problems. The man was not surprised Spy when he found several bugs in the corner of his room. Rumor had it that, for years, the government building had been plagued with problems. The man was not surprised Curtain when he found several bugs in the corner of his room. Measure response times: Is it a Swinney et al. (1979) word?  Testing Lexical Processing: Extending Lexical Decision Tasks 1600 Measure response times: Is it a 1200 word? Reaction Time (ms) 800 What is the pattern in the data? 400 0 Ant Spy Curtain What does this data tell us about Word lexical representations? Swinney et al. (1979)  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #2: How do we match words to meanings? Define lexical processing Role of context and frequency Lexical decision tasks  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #3: How do we give structure to strings of words (ie sentences)?  Parsing: What is it? Parsing: Decompose language into parts Consider/identify grammar (syntax) of a sentence to make sense of it Garden-path model: There are “preferred interpretations” of sentences We first interpret sentences in this preferred interpretation Update our interpretation if conflicting syntactic information comes in Syntax-first approach Gazzaniga et al., 2009  Parsing: Principles of the Garden-Path Model Minimal attachment: Fewest number of nodes combining to make a grammatical sentence = first interpretation Less computationally expensive Gazzaniga et al., 2009  Parsing: Principles of the Garden-Path Model The man who whistles tunes pianos Principle of late closure: Attach new words to the phrase we have already heard/are currently processing ‘The man who whistles tunes’ -> ‘tunes’ = noun (whistle a tune) ’The man who whistles tunes PIANOS’ -> ‘tunes’ = verb (tune a piano) Violation of late closure -> re-assign ‘tunes’ to verb Take longer to process sentences Gazzaniga et al., 2009  Parsing: Sometimes Semantics > Garden-Path! Methods: -Give semantically-possible or -impossible sentences to participants: (1) The defendant examined by the lawyer turned out to be unreliable. (2) The evidence examined by the lawyer turned out to be unreliable Results: Reading times GREATER for sentence (1) than sentence (2) Especially for the phrase “by the lawyer” Principle of Late Closure NOT APPLIED to sentence (2) Trueswell et al. (1994)  Language comprehension: How do we do it? Three big requirements for speech comprehension: 1) Phonological processing 2) Lexical processing 3) Parsing Big Q #3: How do we give structure to strings of words (ie sentences)? Garden-path model 2 principles of the garden-path model Role of semantics  Write down the ‘muddiest point’ for you about Language Comprehension What didn’t you understand? What outstanding questions do you have about the material so far? Share your ‘muddiest point’ with a partner – can you find answers to your ‘muddiest point’?  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Learning Outcomes for Rest of Slides Answer the question using data: Does language influence perception? Describe key language disorders and their neurobiology Identify shared dimensions of music and language Explain the SSIRH for music and language Answer the question using data: Does language influence music perception?  Does language influence cognition more generally? Linguistic relativity: Language acts as a filter on cognition Language affects cognition Linguistic universalism: Differences between languages are superficial No big effects of language on cognition  Is there support for linguistic relativity? Sapir-Whorf hypothesis The language we speak influences the world we perceive and how the mind is organised Mixed evidence for this hypothesis Russian English Measure: Reaction time Winawer et al. (2007)  Is there support for linguistic relativity? Cross-colour category Within-colour category Russian English Faster reaction times to two different blue categories for Russian speakers, but not for English speakers Interpretation: Effect of language on category formation and perception  Language disorders Aphasia: The disruption of language caused by a brain-related disorder Broca’s Aphasia: production (expressive aphasia) Wernicke’s Aphasia: comprehension (receptive aphasia)  Expressive Aphasia “Son…university Broca’s Aphasia …smart…good… good…” severe effort to produce speech short (one-word) utterances speech is agrammatical comprehension is relatively unimpaired Written or spoken Complex grammar comprehension impaired  Expressive Aphasia Early understanding: Damage to Broca’s area sufficient to cause expressive aphasia Current understanding: Damage to deeper brain structures involved in expressive aphasia  Receptive Aphasia “I called my mother on the television and did not Wernicke’s Aphasia understand the door. My romer is tomorrow morning, I invented words/inappropriate words think.” syntactic aspects of speech is preserved speech production is fluid but nonsensical comprehension is severely impaired reading & writing impaired  Receptive Aphasia Early understanding: Damage to Wernicke’s area sufficient to cause receptive aphasia Current understanding: Damage to deeper brain structures involved in receptive aphasia  Language is supported by a network Arcuate fasciculus: White matter tract that connects posterior language areas (Wernicke’s area) to anterior language areas (Broca’s area) Move away from Functional Localisation of speech/language areas  Outline Basics of language Major views on language acquisition (ie. Learning) How do we understand language? Focus on comprehension Language and cognition Language disorders The language-music connection  Music and Language (1) Do language and music share processing resources? (2) How does language influence music perception?  Two perspectives on music and language Domain-general abilities and resources Domain-specific abilities and resources Big question: Do music and language share resources? If so, which resources?  What is shared between music and speech? Shared acoustic properties 1) Rhythm – hierarchical in music and speech Music Speech From Fiveash et al. (2021)  What is shared between music and speech? Shared acoustic properties 2) Pitch – convey information Melody Intonation What is shared between music and language? Full language Auditory but… Organised symbols mapped to meaning  What is shared between music and language? Structure: Syntactic Organisation Rules guide how elements of music and language are combined Music: tone -> chord -> key Language: phonemes -> words -> phrases -> sentences Core idea: Maybe music and language share syntactic processing resources  What is shared between music and language? Structure: Syntactic Organisation of Music Tones -> chords Which combination of tones sounds ‘right’ to you?  What is shared between music and language? Structure: Syntactic Organisation of Music Chords -> Sequences (keys) Which sequence of chords sounds ‘right’ to you? from Patel (2003)  What is shared between music and language? Structure: Syntactic Organisation of Music Tonal Pitch Space Theory (TPST): How we perceive pitch structure and pitch relationships Hierarchical The combination of some pitches is more ‘stable’ than others C major scale: C D E F G A B Not all tones in the scale are as perceptually ‘stable’ C>G>E… from Patel (2003)  What is shared between music and language? Structure: Syntactic Organisation of Language Dependency Locality Theory (DLT): Explains how we connect words to one another even when they are separated in space/time Example: The meteorologist said it was supposed to snow today, but I guess their prediction was wrong. Integrate words in a sentence More complex sentence structure = more difficult to connect (integrate) words from Patel (2003)  What is shared between music and language? Structure: Syntactic Organisation of Music and Language What do TPST and DLT both say about music and language syntax? Integration must occur between elements in a sequence Greater distances between elements = more difficult connection integration of elements Shared aspect of music and language: Shared resources for syntactic integration Shared Syntactic Integration Resource Hypothesis (SSIRH) from Patel (2003)  Testing shared syntactic processing in Music and Language 1) Easy syntax condition: Hear simultaneous music and language that have expected syntactic structure 2) Hard syntax condition: Hear simultaneous music and language that have unexpected syntactic structure 3) Measure sentence comprehension: Compare easy syntax condition vs hard syntax condition Logic: If language and music share syntactic processing resources, then there should be worse comprehension when music AND language syntax are hard Patel et al. (2009)  What is the important Correct Comprehension of Easy L Syntax Hard L Syntax pattern in this data? Sentences (%) Is there evidence that music and language share syntactic Easy Musical Hard Musical Auditory processing resources? Syntax Syntax Control Patel et al. (2009)  Does the language you speak influence music perception? Different languages have more/less important acoustic features Tonal languages: Pitches carry meaning Pitch-accented languages: Some pitch variety, but not fully tonal Non-tonal languages: Pitches do not carry meaning Do you think the language you speak affects music perception? If so, how? If not, why not? Liu et al., 2023  Testing whether the language you speak influences music perception Melody Discrimination N = ~ 600 000 Mistuned Instrument 54 languages Mandarin, Thai … Croatian … English, Tagalog … Beat Alignment Liu et al., 2023  Testing whether the language you speak influences music perception Melody Discrimination Mistuned Instrument Beat Alignment Big Size of statistical effect effect Small effect Liu et al., 2023  Music and Language: Recap Music and language share certain features Acoustic Structural SSIRH: Music and language share syntactic processing resources The language you speak can influence music perception! Write down the ‘muddiest point’ for you about Music and Language What didn’t you understand? What outstanding questions do you have about the material so far? In-class assignment: Post one of your muddiest points on the Discussion Board on Blackboard (W12-MP) Can be something you did not get an answer to (good idea!) Can be something you DID get an answer to (post both the Q and the A -> could help someone else) Bonus assignment (up to.5%): (Correctly) answer someone’s muddiest point Q on Blackboard Add a helpful example to a muddiest point Q/A Correct a muddiest point A Due by Dec 4 11:30 am  Next week: Final week! Infant-directed speech and song No textbook reading: Podcast (Hidden Brain…) plus journal article on Blackboard Short final exam review