Multiword speech

When do infants put words together to create multiword utterances/ multiword speech

18 months- 2 years

Broad theoretical approaches to multi word speech + THEIR SYNONYMS

• nativist → generativist

• constructivist → usage-based

Syntax and its benefits

Way in which a language allows words to be combined/ arranged

1. enables understanding between speakers, e.g. who did what to whom

2. allows productivity: with a finite set of words we can produce an infinite no. of possible sentences

grammatic category of words → grammatical role of ppts → meaning

noun phrase → subject → agent

noun phrase → object → patient

verb → action

2 features about language

• species-specific: little evidence that other primates can acquire syntax even with intensive training

• species-universal: all children acquire majority of grammar of their language by 5 years.

Characteristics of early word combinations

• mainly content words, e.g. nouns, verbs, adjectives

• here-and-now: easily understood in context

• creative: e.g. more sing → trying to ask for more singing

• observes adult word order: truck gone rather than gone truck

suggests organising principles

Nativist view regarding rules

Rules are item-specific: based on individual words/schemas, e.g. get + X, child knows that X can be a variety of words.

Constructivist view regarding views

Rules are abstract and based on grammatical categories, e.g. verb + object OR subject = verb. So rules aren’t restricted, allowing all utterances possible in adult language

Constructivist approach

Grammar is used for communication and infants are motivated to learn to communicate.

How can grammar be learned

Using general cognitive learning mechanisms:

• communicative intention-reading

• drawing analogies (bringing similarities together)

• distributional learning (pattern finding, which bits of language go together)

Evidence that supports the constructivist approach

• children with lexically-based linguistic representations

• high frequency items are learned early

• only gradual generalisation across exemplars to create more abstract syntactic categories + rules

The verb island hypothesis: Tomasello

Evidence to support constructivist approach

• suggests that verbs are the most important item children hear NOT grammar

• children do not initially understand abstract syntactic categories, like nouns and verbs, instead they form islands around each verb

Experimental evidence for verb island hypothesis: Akhtar + Tomasello

With unfamiliar/ novel verbs, children under 3 cannot explain who is doing what to who, but can do so with familiar words. Shows that they don’t have a general understanding of syntactic structures and can’t generalise the use of words.

Limited (lexical) constructions

Evidence for constructivist approach

• early utterances are based around individual lexical items but not exclusively verbs, e.g. Where’s X gone? X can mean a variety of things

Any high frequency (group of) words can form basis for organisation of child’s linguistic system

How do children link their lexically-based constructions to form more adult-like grammar

• structure combining

• semantic analogy

• distributional learning

Structure combining: Lieven et al., 2003

3 operations: substitution, addition + drop

Complex utterances across 6 weeks are based around repetitions/ small changes to what they’ve said before.

Most changes involve simple substitutions within a lexically-based frame

Suggests that children operate with an inventory of SPECIFIC UTTERANCES + limited mechanisms for altering these to match context

Semantic analogy: Matthews + Bannard

Children learn a no. of words before they recognise similarities between them + build general schemas.

Children made less errors repeating sequences when words in 4th slot were more similar. Suggests overlap in meanings helps build flexible constructions

Distributional learning: Childers + Tomasello

Ability to learn co-occurrence of characteristics of the input, i.e. which words occur together or similar contexts, e.g. verbs: -ing

Pronouns helped children extract a more abstract representation of S-V-O sentence structure when novel words were used. VS noun phrase only condition

Critical evaluation of studies into multi-word learning

• production studies are difficult for children: significant memory load in remembering/ recalling novel words etc.

• production studies may underestimate how abstract children’s knowledge of sentence structure really is

• not specified how sentence structures become gradually more abstract over development

Nativist approach assumptions

Assumes that children approach task of learning language with innate machinery that is specific to language.

• grammar is a symbolic computational system which processes relationships between abstract variables

• grammatical categories + rules are given apriori in child’s brain from birth

• all-or-nothing quality

Name for the innate machinery specific to language

Language Acquisition Device or Universal Grammar

Difference between nativists and constructivists about why they believe children’s utterances are creative

nativist → children have access to innate grammatical rules

constructivist → creativity based on use of lexical frames learned from language they hear

Difference between nativists and constructivists about adult word order

nativist → adult word order has an abstract rule: S-V-O

constructivist → observe adult word order because they pick up highly frequent lexical frames from input, e.g. I- want- a drink

Difference between nativists and constructivists about generalisations

nativists → adding inflections like -s is evidence of abstract (innate) rules

constructivists → children learn these patterns gradually from distributional analysis of language they hear

Principles and parameters of UG

• all possible rules for languages are innate

• grammar is universal: rules of grammar apply in all languages

• when rules do differ across languages, it’s in a highly constrained way which is encoded by parameters

• children need to work out which parameter settings apply for language they’re learning

Examples of parameter settings (2)

• word order: verb-object (English) or object-verb (Japanese)

• subject use: some languages make subjects obligatory (English) but others make it optional (Spanish)

What are some theoretical advantages of UG (2)

• avoids problem of explaining how children acquire complex grammatical rules

• allows a unified theory of acquisition across languages whilst also explaining how they differ

Empirical evidence for principles + parameters

1. early utterances usually observe adult word order: taken as evidence the relevant parameter is set

2. children are productive from early on, e.g. all gone sticky- evidence they’re applying rules of grammar

3. preferential looking studies: evidence that children understand role of S-V-O construction before 2 and under

Theoretical problems for UG

1. parameters aren’t clearly specified, e.g. how many are there? which aspects of language do they code for?

2. unclear how children avoid setting parameters incorrectly, e.g. want a drink

3. bilingualism: how do children have 2+ versions of same parameters

Continuity account

Idea that language develops overtime. Limitations seen are from performance NOT limited knowledge

Radford’s maturational model (1990)

Lexical stage (20mo): utterances consist of mainly content words

Functional stage (24mo): child’s innate grammar matures and parts the govern complex grammatical components are switched on, e.g. auxiliary verbs, determines, inflections

Advantages of Radford’s maturational model

• explains why early utterances are not fully grammatical

• allows for development overtime: likely to fit empirical data

• similar trajectory for all children, e.g. deaf, blind, normal hearing despite their experiences being different

Problems with Radford’s maturational model

• difficult to identify precise point in development when these components are switched on

• children show some use of complex grammatical functions before 24 months (functional stage)

• at 24 months, use of functional words are related to lexical frames

Linking problem

Questions how children link up their innate knowledge of grammatical categories to the words they’re hearing.

UG is described as having innate knowledge of nouns, verbs etc. but caregivers don’t label particular words as nouns, verbs etc.

Proposed solution to the linking problem: Pinker

SEMANTIC BOOTSTRAPPING

• assumes grammatical categories and rules are innate, and children use semantics to map words in the input onto these syntactic categories by using innate linking rules to map semantics onto syntax.

Linking semantic roles to syntactic roles

agent= subject

patient= object

Problem with semantic bootstrapping and a solution to this

Not always this simple:

• not all verbs are actions, e.g. believe, need

• not all nouns are concrete, e.g. idea, dream

• not all subjects are agents, e.g. she wants a drink

Use distributional analysis to determine word order for language from prototypical sentences. Then apply knowledge of word order to work out grammatical category of more abstract terms

Advantages of semantic bootstrapping

• explains how children break into innate system

• explains why early utterances follow adult word order

• explains how children learn verbs which are not actions or nouns which aren’t objects

Other problems for semantic bootstrapping

• many early lexically-specific utterances are not semantically prototypical, so unlikely to be based on innate knowledge of linking semantic rules

• passive sentences: object becomes subject, semantic bootstrapping would cause child to conclude that language is object-verb-subject

• also children use passive sentences from early on, so can’t say that these parameter doesn’t mature until 5