Early Multi-word Speech: Constructivist Approaches PDF

Early multi-word speech: Constructivist approaches Professor Anna Theakston PSYC21021: Topics in Developmental Psychology  Caregiver-child interaction Overview So far, focus has been...

Early multi-word speech: Constructivist approaches Professor Anna Theakston PSYC21021: Topics in Developmental Psychology  Caregiver-child interaction Overview So far, focus has been on learning to communicate and early words. Now, how do children put words together to create multiword utterances/multiword speech. Typically happens between 18-months to 2 years of age. Two broad theoretical approaches; nativist (or generativist) accounts and constructivist (or usage- based) accounts. This week, constructivist accounts, next week nativist accounts words that adult uses. reports Listen to a 2-yr-old (~line 750) Learning Objectives By the end of this week you should be able to: Describe the nature of children's early multiword combinations Outline two rule systems that have been put forward to explain children's multiword speech Explain the key assumptions of the constructivist (usage- based) account Evaluate constructivist (usage-based) accounts of language acquisition using appropriate empirical evidence Lecture Structure Part A: Background Part B: Constructivist approaches – theory & initial evidence Part C: Building an adult grammar Overall summary & areas for evaluation Part A: Background crammer 1. What is syntax? 2. What needs explaining? 3. What are early multiword combinations like? 4. What kinds of rules could underpin early multiword utterances  grammer. 1. What is syntax? The ways in which a language allows words to be combined: Enables understanding between speakers, e.g. ‘who did what to whom’ Allows productivity – with a finite set of words we can produce an infinite number of possible sentences. The language of syntax Grammatical category of words: Noun Phrase Verb Noun Phrase ↓ Grammatical role of participants: ↓ ↓ Subject Object Sentence: Mary feeds the rabbits Meaning: AGENT ACTION PATIENT ↳ doing ↳ the ↳ affected by the action action. the action 2. What needs explaining? Language is: Species-specific: Little evidence other primates can acquire syntax even with intensive training. Species-universal: virtually all children have acquired the majority of the grammar of their language by 5yrs. Kanzi the bonobo with E. Sue Savage – Rumbaugh using lexigrams on portable keyboard to communicate (Copyright © 2006 Great Ape Trust; courtesy of Duane Rumbaugh) 3. What are early word combinations like? contends - Daddy shirt, More milk, Want drink, Got dirty hand are happines  Mainly content words t thingsthat them.  Refers to here-and-now, easily understood in context.  Creative More sing, All gone sticky, other one spoon > add - together words , not  Observes adult word order truck gone vs. gone truck just pure repetition. Suggests some kind of organising principles.  4. Lexical (word-based) rules? lexical based. - or Rules item-specific - based on individual words or schemas (sets of words). have something to > - patter it of Get + X Get -truck/drink/crayon… Where’s the + X Where’s the truck/man/car… - truck , man , car ↳ X Limited variety of utterances until children are able to generalise between schemas ↳ sets of words. e.g. Lieven et al, 1997; Pine et al, 1998 refers to a group of words that can be used in that * = S10t. Syntactic (grammatical) rules? Rules abstract – based on grammatical categories. Verb+Object Get truck, Causes problems… Subject+Verb She laughs, It’s raining, Confident people succeed… Rules not restricted, therefore allow all utterances possible in the adult language (e.g. Valian, 1991; Radford, 1990) Interim summary Children’s early multiword utterances are not random, nor simply imitations of what they have heard Children learn language, but other species do not demonstrate the same impressive abilities. So how are children able to put words together into sentences, what kind of knowledge do they need to be able to do this, and where does it come from? Part B: Constructivist approaches: theory & initial evidence 1. What is the Constructivist approach? 2. What kind of evidence would support the approach? 3. Evidence for lexically-based (word-based) learning 1. What is the constructivist approach (usage-based) Grammar is used for communication > - babbling. Infants are motivated to learn to communicate Grammar can be learned using general cognitive learning mechanisms Communicative intention-reading Drawing analogies Distributional learning  BBC Tiny Happy People – mealtimes routine The role of routines utterances that make them know what will happen. next repeated Routines allow children to predict what happens next and therefore what the language they are hearing might refer to. Repetitive chunks of language can then be learned in context where the relation between linguistic form and meaning is more transparent. Book reading routine 2. What kind of evidence would support a constructivist/ usage-based approach? Children begin with lexically-based linguistic representations High frequency items are learned early Only gradual generalisation across exemplars to create more abstract syntactic categories and rules 3. Evidence: The verb island hypothesis Knowledge of grammar tied to individual verbs until 2½- 3yrs. Child initially unable to generalise between verbs with similar meanings or used in similar sentence types. FORM [X KICK Y] Kicker > - being MEANING [KICKER KICKEE] ↳ kicks Tomasello, 1992  Experimental evidence With familiar verbs (e.g. chasing), 2-yr-olds able to describe actions correctly to explain who is chasing, and F whom is being chased. have verb island · Perhaps Children a able to order because they are only the words aroundwhenthey knod & But with unfamiliar (novel) verbs (e.g. weefing), before 3yrs children struggle to explain who is doing what to whom. 4 before Why? 3 years old , they cannot order words aroundwhenan win the verb. (e.g. Akhtar & Tomasello, 1997) Evidence: Limited (lexical) constructions See activity on Blackboard Argue children’s early utterances based around individual lexical items (words) but not exclusively verbs I+X Can I + X Where’s X gone? X + go More + X Don’t + X ‘X’ represents a set of possible words used in the slot Any high frequency word / group of words can form the basis for organisation of the child’s linguistic system. The constructions children learn reflect the frequency of particular patterns in the input. (e.g. Lieven et al., 1997; Pine et al., 1998) Interim Summary Good evidence that children’s early utterances are more restricted than those of adults. How do children start to ‘join up’ the different parts of their developing linguistic knowledge? Part C: Building an adult-like grammar How do children link up their lexically-based constructions to form a more adult-like grammar? 1. Structure combining 2. Semantic analogy 3. Distributional learning 1. Structure combining (Lieven, et al, 2003) How do children’s utterances build on what they have previously said? recorded all utterances child said for weeks. > - Dense diary study of a single child for 6 weeks at 2;0 Recorded for 5 hours/week, and written diary of all new utterances kept by mother. All utterances on last hour-long recording noted – ‘Target’. - All previous recordings searched for closest match – ‘Source’. - Method  Identify:  What changes required to change closest matching utterance - the ‘source’ - into the ‘target’ utterance (operations) Target utterance Source utterance Operation -> substituted  I got the butter I got the door (substitution)  It’s burning here It’s burning (addition) word ↳ additional  And horse And a horse (drop) Results Yesterday Last week 295 multiword utterances My did it  186 repetitions (63%)  158 repetitions of something child said previously  28 immediate repetitions of mother  109 novel utterances (37%, of these ¾ single operation change)  68 substitutions + 12 add on + 1 drop  22 utterances required 2 operations, e.g. add on + substitution Where’s Daddy’s work? Where’s my Daddy’s cup of tea  6 utterances required 3 or more operations (substitute, drop, add) I can’t put it back on I don’t put it (___) on there Conclusions  Many of the child’s apparently complex utterances are based around repetitions or small changes to what she has said before.  Most changes involve simple substitutions within a lexically-based frame, or the addition or subtraction of a single word.  Suggests child is operating with an extensive inventory of specific utterances, and fairly limited mechanisms for altering these utterances to match the demands of the discourse context.  2. Semantic analogy Children need to learn a number of verbs before they can recognise similarities between them and begin to build more general schemas. Commonalities reinforced, differences forgotten -set of words. X pull Y Gene wish read X push Y ↳ X = “Doer” X VERB Y a pusheide X = “Doer” X kick Y X roll Y ACTION Y = “done-to” Evidence: repeating sequences (Matthews & Bannard, 2010) 2 & 3-yr-olds asked to repeat 4-word sequences [FRAME] [SLOT] relates T Back in the ’box/case/town’ (higher similarity) It’s time for ‘lunch/soup/drums‘ (lower similarity) - not related Manipulated 3-word frame by similarity of meaning of items in 4th ‘slot’ Children made fewer errors when items that normally occur in the slot are more similar – suggests overlap in meaning helps build flexible constructions 3. Distributional learning The ability to learn the co-occurrence characteristics of the input, i.e. which words occur together or in similar contexts. Verb-ing/-ed/s walking/walked/walks; jumping/jumped/jumps Noun-s/’s dogs/dog’s; cats/cat’s -pluraform Why are the tigers eating? / Why- -wal are the boys singing? Why is the dog running? / Why -- is the girl dancing? ↳ singular form Jingular. Experimental evidence for distributional learning 2-yr-olds exposed to multiple transitive sentences of form X is Verb-ing Y with familiar verbs Noun Phrase only condition – all Xs and Ys are lexical nouns nourrace nowrase. The cat is chasing the mouse The bear is hugging the fox Mixed condition – Xs and Ys are combination of lexical nouns and pronouns The cat is chasing the mouse / He is chasing him The bear is hugging the fox / He is hugging him ours Lexical pronous (Childers & Tomasello, 1998) Experimental evidence for distributional learning Children taught novel verb (This is called dacking) to describe a new action between two participants. Asked What’s happening here? to elicit description In which condition are children more likely to show generalisation of the X is V-ing Y lexical frame? i.e. to say ‘The dog/he is dacking the lion/him’ [vs. just repeating novel word, i.e. ‘dacking’, or only producing partial utterance mentioning just one participant, i.e. ‘dacking the lion/him’ or ‘the dog/he’s dacking’ ] (Childers & Tomasello, 1998)  Experimental evidence for distributional learning nowed result. S a ‘This is called dacking, What’s 90 mixed happening here?’ 80 70 60 Control 50 NP only Pronouns helped children 40 Mixed 30 extract a more abstract 20 10 representation of the Subject- 0 Verb-Object sentence structure for use with novel (unfamiliar) verbs for structure with - Pronouns benefitted children sentence novel (unfamiliar) verbs. % children producing ‘X is (Childers & Tomasello, 1998) dacking Y’ responses Interim Summary Studies of children’s language production suggest early language not organised around same categories and rules as used by adult speakers. Evidence for gradual generalisations based on similarities in form and meaning of sentences. Overall Summary Children begin to combine words together at 18-24 months Constructivist theorists argue that children access meaning and learn to combine words by interpreting the intentions of their interlocutors – from hearing language used in predictable contexts Children build up grammar by starting with more limited scope rules (e.g. lexical rules) than those used by adults and using general cognitive mechanisms to generalise Critical Evaluation Production studies are difficult for children – significant memory load in remembering and recalling novel words, planning entire sentences. Do production studies underestimate how abstract children’s knowledge of sentence structure really is? Exactly how sentence structures become gradually more abstract over development is not clearly specified.  Reading Guidance The material covered in this lecture will be assessed via a short answer & MCQ Exam. Note, questions may require you to reason about and make inferences from the information you have learned. Students who have a better understanding of the material are better able to manipulate the information to arrive at the correct answers. You might find it helpful to read sections from: Clark, E. (2009/2016). First Language Acquisition. The 2nd Edition is available as an electronic VLeBook via the Library website that you can download or read online, and is suitable for this course unit. The updated 3rd Edition is available in hard copy from the library. For this lecture content, the following sections are most helpful: Part II, Chapter 7, (First combinations, first constructions), pp. 157-162; pp. 167-174 If you are interested to learn more about how children might 'scale up' their early constructions towards a more adult-like grammar, Clark's Chapter 9 on 'Adding complexity within clauses' is worth a look, although it covers a wider range of material and in more depth than we have considered in this lecture.