Language Acquisition JLP285H5 Fall 2024 PDF

LANGUAGE ACQUISITION JLP285H5, Fall 2024 Professor Atkinson  2 Today How is language acquired?  3 How do we study language acquisition?  4 Examples of...

LANGUAGE ACQUISITION JLP285H5, Fall 2024 Professor Atkinson  2 Today How is language acquired?  3 How do we study language acquisition?  4 Examples of Methodologies Diary studies → observations of behaviour, usually of a single subject (often the researcher’s own child)  5 Diary Studies: Charles Darwin (1877) The Expression of the emotions in man and Animals (book published in 1872) ‘A biographical sketch of an infant’ (article published in Mind in 1877) Based on notes of his own children’s development (mostly of his first son, William Erasmus)  6 Diary Studies: Charles Darwin (1877) “When exactly seven months old, he made the great step of associating his nurse with her name, so that if I called it out he would look round for her.” “During the next four months the former infant associated many things and actions with words; thus when asked for a kiss he would protrude his lips and keep still” “When 46 days old, he first made little noises without any meaning to please himself, and these soon became varied.”  7 Diary Studies: Charles Darwin (1877) “At exactly the age of a year, he made the great step of inventing a word for food, namely, mum, but what led him to it I did not discover. […] But he also used mum as a substantive of wide signification; thus he called sugar shu-mum, and a little later after he had learned the word ‘black,’ he called liquorice black-shu-mum,– black-sugar-food.” “Before he was a year old, he understood intonations and gestures, as well as several words and short sentences.”  8 Diary Studies in the 21st Century Deb Roy (MIT, 2011) – big data version of a “diary” study https://youtu.be/RE4ce4mexrU  9 Examples of Methodologies Diary studies → observations of behavior, usually of a single subject (often the researcher’s own child) Corpus analysis → Child Language Data Exchange System (CHILDES)  10 CHILDES Database of child language transcriptions Gathered & maintained by Brian MacWhinney at CMU https://childes.talkbank.org/  11 Examples of Methodologies Diary studies → observations of behavior, usually of a single subject (often the researcher’s own child) Corpus analysis → Child Language Data Exchange System (CHILDES) Psycholinguistic experiments → studies designed to test children’s knowledge or processing of language Ability to perceive / understand often ahead of ability to generate / produce!  12 Psycholinguistic Methodologies Preferential Looking High Amplitude Sucking Head Turn Preference (LSCP Infant Lab) Conditioned Head Turn (Werker et al. 1998)  13 High Amplitude Sucking Procedure Pacifier contains sensor to monitor sucking rates Each time infant sucks pacifier, a stimulus is played LSCP Infant Lab  14 High Amplitude Sucking Procedure Initially sucking rate increases (novelty) then decreases Decline in response = habituation LSCP Infant Lab  15 High Amplitude Sucking Procedure When sucking rate declines to a set rate (habituation criterion), auditory stimulus is changed LSCP Infant Lab  16 High Amplitude Sucking Procedure If sucking rate increases after the change, then we know the infant has detected the change Renewed response = dishabituation LSCP Infant Lab  17 Developmental Methodologies High Amplitude Sucking only works with infants from birth to 4- months-old The Conditioned Head Turn Paradigm can be used for 6- to 12- month-old infants Age range = before & around the onset of word learning  18 Conditioned Head Turn Paradigm https://youtu.be/CSMjKDZvNWA  19 Conditioned Head Turn Paradigm Toy that lights up & moves at the experimenter’s command Werker et al. (1998)  20 Conditioning Phase Infant trained with 2 clearly different auditory stimuli (bell & whistle) 1. Each time the sound changes, the toy is activated 2. Toy is activated only if the sounds changes AND the infant turns to look at it  21 Preferential Listening Procedure Head Turn Preference  22 Preferential Listening Procedure Head Turn Preference (e.g., Jusczyk & Aslin 1995) / Preferential Listening Sounds are played through a large TV monitor with speakers Measure how long infants orient their gaze / head toward the monitor → when bored, they’ll look away If gaze duration is significantly longer in one condition than in another, indicates discrimination of stimuli  23 Truth Value Judgment Task (TVJT) When you know what a sentence means, you know what the world would have to be like for the sentence to be true To know the meaning of a sentence = to know its truth conditions Truth value judgment tasks measure children’s responses to questions about the truth conditions of sentences  24 Truth Values Declarative sentences have a truth value (True or False) Knowledge about truth conditions + relevant knowledge about the world = you can judge whether a sentence is true or false E.g., Dr. Atkinson speaks English = TRUE  25 TVJT Developed & popularized by Stephen Crain in the late 1990s Experimenter acts out story in front of the child using toys & props After the story, another experimenter manipulates a puppet (traditionally Kermit the Frog) & presents a test sentence to the child Child asked to say if Kermit was right or wrong  26 Truth Value Judgment Task “I know what happened in this story…”  27 Sounds to words: Speech Production  28 Prelinguistic Sounds 0;0 – 0;3 → cries / coughs / burps / wheezes ~0;3 → babbling begins Babbling = repetition of speech-like sounds e.g., ‘ba ba ba’  29 Stages of Babbling 0;2 – 0;3 → first vowel (V) vocalizations Consonant sounds produced in the back of the mouth 0;6 – 0;7 → appearance of canonical babbling Consonant-vowel (CV) syllables: no meaning attached, but resemble adult syllables Consonants produced with lips or front of tongue [m] / [b] / [d] → [baba], [di]  30 Example of Early Babbling 7-month-old babbling: https://youtu.be/3UCK4XCrvoc?si=IV7J6AiFDg0o8KYR  31 Stages of Babbling 0;6 – 1;0 → expand consonant sounds to include stops / nasals / glides ~ 1;0 → production of first word(s) Words & babbling co-exist for several more months  32 Why Babble? Feedback loop = infant can hear their own babbling Does the sound I made correspond with the sound I meant to make? Establish link between articulatory movements (e.g., tongue position, lip position, etc.) & acoustic signal Babbled syllables are 1st step to adult-like words Canonical babbling = CV syllables 1st words resemble babbling Use same consonants & syllable types  33 First Words Period onset production meaningful speech to 50-word vocabulary Ends around 18-months Characterized by simple syllabic structures & small group of consonants & vowels Syllable structure = CV (go) / CVC (sit) / CVCV (baby) Consonants = stops / nasals / glides Notably absent → infrequent sounds (e.g., [v]), those that require articulatory precision (e.g., [ʃ], [ʒ], [ɹ], [θ], [ð])  34 Cross-Linguistic First Words Similar properties across languages CV syllables predominate Stops / nasals / glides are frequent Language-specific features → sound classes / syllable types / stress patterns English: more CVC (ball) / two-syllable with stress on first syllable (mómmy, bóttle) French: more two-syllable words / stress on last syllable / nasal consonants  35 Overextension Word meanings are abstract Must be extended to all members of a class Apple!  36 Overextension Children overextend word meanings to related referents → usually some kind of conceptual overlap Word use “beyond their conventional range of meaning” (CL, p. 151) 1;0 1;2 “Papa” (Hildegard: Leopold 1939)  37 Overextension Very common between 1;6 & 2;6 Up to 40% of child word uses Other examples… “car” → car, skateboard, bike, wagon “fly” (noun) → fly, specks of dirt, dust, all small insects, child’s own toes, crumbs of bread, toad “baby” → self, other children, any child, doll “ticktock” → square-faced watch, clock, all clocks & watches, gas meter, fire hose on spool, bath scale w/ round dial, round eraser  38 Sounds to words: Speech Perception  39 Segmentation Problem Acoustic input is not physically segmented Illusion of silence  there are no silent gaps in the wave form, even though we may “hear” some  40 Lack of Invariance No one-to-one mapping from acoustic signal to linguistic representation 1st [I] in ‘cognitive’ different from 2nd [I] Same for [n] in ‘cognitive’ & [n] in ‘science [n][ɪ] [ɪ] [n]  41 How do phonemic categories develop? Children need to learn these abstract phonological constructs, but they are only exposed to the acoustic data in the in input They don’t have direct access to phonemes (which are in adults’ mental lexicon) So, how can they learn phonemes of their native language? Or do they perhaps have all the phonemes from birth? Approach: Test categorical perception in very young infants  42 Eimas et al. (1971) Can 4-month-olds perceive categorically? BA PA BA1 BA2 BA3 PA1 PA2 PA3 -20 0 20 40 60 80 VOT in milliseconds 3 conditions 20D (cross-category): {+20 & +40} 20S (within-category): {0 & +20 | +60 & +80} Control: randomly assigned, no change LSCP Infant Lab  43 Eimas et al. (1971) Cross- Within- Control category category  44 Categorical Perception in Infants 4-month-olds & even 1-month-olds categorically perceive [b] & [p] (Eimas et al. 1971) Newborns are sensitive to many (if not all) phonemic contrasts tested Infants are born to be (mostly) universal listeners? Not a specifically human ability → chinchillas can do it too!  45 Universal Phonetic Space Japanese English French Universal Phonetic Inventory  46 Newborns are Universal Listeners Newborns can distinguish all possible phonemes in all languages Adults have difficulty discriminating speech sounds that are not contrastive in their native language Janet Werker  47 Longitudinal Study Werker (1995)  48 Using Phonemic Information Werker’s studies show that 10-month-olds lose sensitivity to non-native phonemic categories Universal (i.e., innate) listener → selective listener? Language learning as selection process  49 Finding word meaning  50 Learning Word Meanings Gavagai Problem Infinite mapping possibilities between a word label & reference How do children learn words so early & quickly??  51 Word Learning in Real Life (Medina et al. 2011)  52 Solving the Gavagai Problem Children equipped with a number of learning biases Pay attention to certain aspects of scene = more likely to make same assumptions about words as adults Disambiguate (i.e., select among word meaning hypotheses) using cues & constraints Role of (visual) perception Conceptual bias Syntactic information Etc.  53 Word Learning Biases Bias Description Whole Object New words name whole objects, not parts or properties Shape Objects that have the same shape will have the same name Mutual Words are mutually exclusive: each object will Exclusivity have 1 & only 1 label (Based on Table 6.3)  54 Syntactic Cues This is a zav.  55 Syntactic Cues Give me the zav. Which one?  56 Syntactic Cues This is a zav one.  57 Syntactic Cues Give me the zav one. Which one?  58 Syntactic Bootstrapping Children generalize word meaning on the basis of: Color if the novel word is used in Adj context Shape if the novel word is used in N context Syntactic bootstrapping = using syntax to learn word meaning Syntactic environments constrain the hypothesis space of possible meanings (Landau & Gleitman 1985; Waxman 1999; Waxman & Markow 1998)  59 Look, there’s blicking!  60 Look, she’s blicking it!  61 Syntactic Bootstrapping & Verb Learning Verb meanings & the number of syntactic arguments tend to correlate Causative (e.g., X bends Y) = 2 overt arguments (subj & obj) Non-causative (e.g., X sleeps) = 1 argument (subj) Transfer of object (e.g., X sends Y to Z) = 3 arguments (subj, direct object, indirect object) She’s blicking it! 2-participant event → causative Referential disambiguation (at least)  62 Syntactic Bootstrapping 28-month-olds just hearing sentences (with no scene / reference) (Yuan & Fisher 2009)  63 Syntactic Bootstrapping (Yuan & Fisher 2009)  64 Syntactic Bootstrapping (Yuan & Fisher 2009)  65 Summary So Far Language learning: rich hypothesis space → reduction of the hypothesis space All potential phonemes → phonemes of target language Word meaning = narrowing down meaning candidates Learning mechanisms sensitive to certain types of cues that reduce the hypothesis space Syntactic cues constrain the space of possible meanings for novel words  66 Development of Phrase Structure Rules  67 Phrase Structure Abstract Any nouns can go into the NP slots representation Any verbs can go into the V slot Possible to reliably determine which NP is likely to correspond to Agent & which to Patient Test: what if you are presented with a novel V, “gorp”?  68 The duck is gorping the bunny! Test if English-speaking infants can assign a correct Agent-Patient interpretation to SVO sentences with novel verbs Preferential looking method Do 21-month-olds have abstract syntactic structures? (Gertner, Fisher, & Eisengart 2006)  69 The duck is gorping the bunny! https://youtu.be/-p_krUd2CSA (Gertner, Fisher, & Eisengart 2006)  70 Preferential Looking: Test Procedure If children have abstract phrase structure representations, what kind of data would you predict? (Gertner, Fisher, & Eisengart 2006)  71 Results (Gertner, Fisher, & Eisengart 2006)  72 Rule learning & regularization bias  73 Mental Grammar Innate component + learning component Rule system → how do children learn rules? Or do children have rules at all? Could they simply imitate / memorize what they hear?  74 Children as Rule Makers 1) Children do not simply replicate their input Apply appropriate inflection to nonce words  75 Application of Inflectional Rules Jean Berko (Gleason) (1958) Do children know how to produce the correct form of inflectional morphemes? Wug Test! Wug test = examined whether school-aged children could correctly inflect novels words Why novel words?  76 Wug Test: Plurals “wugs” “gutches” [wʌgz] [gʌʧəz]  77 Wug Test: Past Tense “ricked” “spowed” [rɪkt] [spaʊd]  78 Child Taking the Wug Test https://youtu.be/MgB2iMuEZAA  79 Children as Rule Makers 1) Children do not simply replicate their input Apply appropriate inflection to nonce words 2) Overregularization phenomena  80 U-Shaped Learning: Past Tense Stage 1: walked, played, came, went Stage 2: walked, played, comed, goed, holded Stage 3: walked, played, came, went, held 1) Memorization 2) Rules → overregularization 3) Learn irregular verbs (exceptions)  81 Overregularization Errors Errors in application of regular past tense rule: “Daniel’s broked my mast.” “And he sticked his tongue out at Snow White.” “In the old days, they eated bad things.” “Well, Tippy bited me.” “I drawed a picture of a pirate today.” “He shooted the fish.” “I wish you buyed some for me.” (Saxton 2017: Alex, 4;1-4;9)  82 Development of sentence processing mechanisms  83 Reminder: Tanenhaus et al. (1995) Put the apple on the towel in the box. 1-referent scene 2-referent scene  84 Trueswell et al. (1999) Same experiment as Tanenhaus et al. (1995), but with 5-year-old children Put the frog on the napkin…in the box John Trueswell UPenn  85 1-Referent, Ambiguous Condition Napkin Box (Incorrect (Correct Goal) Goal) “Put the frog on the napkin…”  86 1-Referent, Unambiguous Condition Napkin Box (Incorrect (Correct Goal) Goal) “Put the frog that’s on the napkin…”  87 Adult-Like Incrementality Incorporate verb information to quickly analyze the ambiguous “on the napkin” as the destination Sensitivity to disambiguating information (“…the frog that’s on the napkin) → they have mature linguistic knowledge  88 However… Non-adult-like comprehension behavior 1 2-referent conditions = no early disambiguation Failure to integrate contextual information?  89 However… Non-adult-like comprehension behavior 2 Failure to revise the initial analysis Put the frog on the napkin… …in the box! ~60% of the time children still move the frog to the empty napkin, regardless of the number of frogs (hopping or doubling error)  90 Double-Edged Sword Incrementality allows efficient comprehension, but it may turn out to be incorrect When incorrect, children fail to revise the initial analysis = kindergarten path effects First interpretation = only interpretation? Poor revision abilities do not stop children from being incremental!  91 Conclusion Linguistic structures are abstract & show cross-linguistic variation, but infants can learn them surprisingly fast Phonemes to words Word meanings, inflectional morphology Phrase structure rules Learning biases allow children to use appropriate information in the input to narrow down the hypothesis space  92 If you’re interested in learning more… Take JLP315H5 Language Development By three years of age, children have mastered many of the complexities of human language. How do they do this so rapidly, and with such ease? In this course, you will examine language acquisition from a cognitive perspective. Topics include the acquisition of speech sounds, sentence structure, and conversational abilities, as well as patterns of development in special populations. You will also learn about childhood bilingualism and social aspects of language development. Hands-on experience analyzing recordings of children will be provided. Offered next term: Winter 2025  93 If you’re interested in learning more… Take JLP481H5 Topics in Developmental Psycholinguistics How do children's language comprehension and production abilities differ from adults? What can research on language acquisition tell us about why language looks the way it does? Developmental psycholinguists use experimental techniques to explore a range of topics in the area of child language comprehension and production. Drawing on cutting-edge interdisciplinary research, we will explore contemporary issues and debates in this area. Offered next term: Winter 2025 Taught by me! :) If you’re interested, check the pre-reqs!  94 If you’re interested in learning more… Researchers at UTM who study related topics: Dr. Christina Vanden Bosch der Nederlanden (Psych) – Language, Attention, Music & Audition Lab https://www.utmlamalab.com/ Dr. Elizabeth Johnson (Psych) – Child Language & Speech Studies Lab https://www.classlab.psycholinguistics.ca/ Me! Dr. Emily Atkinson (Ling) – UTM Language Acquisition & Processing Lab https://emilyeatkinson.net/ https://utmlaplab.ca/