Ling 423/640G: Cognitive Linguistics

Ben Bergen

 

Meeting 24: Learning language

November 20, 2008

 

Learning language

 

How do children learn language, and what is the content of their linguistic knowledge?

·          How do they learn the meanings of words?

·          How do children get the constructions they have, what are those constructions like, and how to they become more adult-like?

·          How does language acquisition differ across languages?

 

A usage-based approach

 

Children appear to learn on the basis of attempting to understand and reproduce utterances.

An utterance is "a linguistic act in which one person expresses towards another, within a single intonation contour, a relatively coherent communicative intention in a communicative context." (Tomasello 2000)

 

What is a communicative intention? It's "one person expressing an intention that another person share attention with her to some third entity." (Tomasello 2000)

 

This type of behavior is something specific to homo sapiens.

·          Chimps and other non-human primates don't direct the attention of conspecifics with physical gestures or vocalization.

·          By contrast, human infants by age 9 or 12 months can already:

o     follow gaze direction and pointing gestures of others

o     imitate the actions of others on objects

o     manipulate others' attention by pointing or holding up objects to show them to others

·          In other words, what has evolved in humans is the ability to learn (or innate knowledge) that other people have minds, and that they can manipulate yours and be manipulate.

·          Given this cognitive leap forward, human children have the cognitive capacity to understand actions as intended manipulations of mental states, in contrast with young of presumably every other species on the planet.

 

Utterances are the minimal communicative unit, and these are what children first grab onto in both understanding and production.

·          Granted, many early utterances are single words, but these are (almost) always uttered in the context of an attempted communicative act.

o     Up! Ball! All gone!

·          Children slowly begin to fill out their utterances with the other relevant words as they proceed to the two-word stage and beyond. This is where grammar begins.

 

Learning Grammar

 

Early grammar isn't structured around grammatical categories of abstract syntactic phrases or transformational rules.

·          It's centered around utterance-level constructions that provide the speech act content of an utterance, like I wanna X, Gimme X, More X, It's a X, X gone, X broken, and so on.

·          These early constructions have slots in them that are filled in by a specific set of fillers

·          How can we tell, though, that this is the correct characterization, and that kids aren't using innate grammatical categories and extremely general syntactic rules?

 

Looking at language use is a good place to start.

If kids have knowledge of universal grammatical categories and are learning general syntactic rules, then they as soon as they learn that two words belong to the same category, and demonstrate knowledge of a particular syntactic arrangement for one, the same should apply to the other.

·          E.g. If you know you can say Cut it and Cut it with the scissors, then given that you also know Hit it, you should also be able to say Hit it with the scissors.

·          But this is not what children do. Early on (before 2.5-3 years), each verb acquires its own set of argument structure patterns, regardless of what the child knows about other verbs.

·          The same is true of matrix verbs - kids say I think X, but not I don't think X, He thinks X, I thought X, etc., while with other matrix verbs they do have these patterns, e.g. I don't like X, He likes X, etc.

 

If kids are really accessing abstract linguistic rules, then they should say pretty much anything that conforms to their grammar and that's relevant for them to say.

·          But in fact, young children almost exclusively only say things that they've heard from others, especially if they've previously said them themselves.

·          In one study (Tomasello et al 2001), a 2-year-old English learner was taped one hour per day, five days a week, for six weeks (455 utterances). What was the kid saying?

o     78% of utterances were completely identical to previous of the child's

o     18% were a modification by one bit of a previous utterance (e.g. I have the ball -> I have the butter)

o     4% involved two modifications, e.g. a new slot filler and an added word

o     3 were more complex. Of these, two were imitations of the caregivers previous utterance.

·          It seems like kids are just repeating the same utterances most of the time, and using constructions with new fillers other times. Occasionally, they do something more complex.

·          Note that this is just with 5 hours per week of data!

 

Conclusions

·          Children start with utterances, which are communicative, because that's what they're programmed for

·          They then identify commonalities among these, which simultaneously produces slots where there is variation among the compared utterance constructions.

·          From these, they extract further generalizations.

·          Throughout, it is the communicative use of language that structures it's acquisition.