Dawson Margin Notes On Pinker "Language Acquisition"

Introduction

Language acquisition is a central topic in cognitive science. "Learning a first language is something every child does successfully in a matter of a few years and without the need for formal lessons. With language so close to the core of what it means to be human, it is not surprising that children's acquisition of language has receive do much attention."

The study of language acquisition can answer many fundamental questions: Do children learn language using language specific mechanisms, or is language acquisition part of general intelligence? Is language unique to humans? What is the relation between language and thought? "We shall see that it is virtually impossible to show how children could learn a language unless one assumes that they have a considerable amount of nonlinguistic cognitive machinery in place before they start. ... Language acquisition might be our best hope of finding out how heredity and environment interact." This is because we know lots about the input and output for language acquisition.

The Biology Of Language Acquisition

Basic claim is that human language is made possible by special adaptations in the course of human evolution.

For example, biological specializations include the shape of the vocal tract. There is no evidence that human language is homologous to ape "language". But this doesn't affect the evolutionary account, because we didn't evolve directly from chimps.

Another biological specialization involves left hemisphere brain circuitry. "The brain mechanisms underlying language are not just those allowing us to be smart in general. Strokes often leave adults with catastrophic losses in language, though not necessarily impaired in other aspects of intelligence." There is a double dissociation here too, because you can find severely retarded individuals with normal linguistic ability.

Finally, evidence suggests that biological maturation storngly affects language acquisition. (NB: This is seen in the Gleitman and Newport chapter, which is also required reading for this week of the course.)

The Course Of Language Acquisition

Language acquisition is studied with a variety of methods: natural observation, and experimental methods such as production tasks, comprehension tasks, judgement tasks.

Language acquisition begins with acquiring the sound patterns of the language in the environment. Focus in infancy is control of vocal tract and the extraction of phonetic distinctions. At the age of approximately 1 year, words are understood and uttered. At 18 months, two changes occur: sharp growth in vocabulary, and the use of primitive syntax to produce 2 word sentences (highly similar across cultures). Amazingly, babies appear to comprehend syntax. "Children's output seems to meet up with a bottleneck at the output end. Their two- and three-word utterances look like samples drawn from longer potential sentences expressing a complete and more complicated idea." From 2-3 years there is an explosion of fluent grammatical conversation, in which sentence length and grammatical complexity increase dramatically. Complexity increases because of the use of embedded structure. The rate of language acquisition is not affected by what language is being learned. By 4 years, children have mastered the language.

Explaining Language Acquisition

Basic question: "How do we explain the course of language acquisition in children -- most importantly, their inevitable and early mastery?" Lots of different mechanisms must be at work. Indeed, the phenomenon is so complex that we need a precise framework for what it involves.

Formal learnability theory provides this framework. (NB:It is a formal answer to the computational question about what language acquisition is.) Formal learning theory specifies four things: A class of languages, one of which is to be learned; a language environment (i.e., a sample of sentences from a to-be-learned language), a learning strategy, and a criterion for success.

One thing that formal learnability theory does is lets us show why language learning is so hard. For example, children don't get much negative envidence (NB: In Gold's terms, they are not informant learners). "Without negative evidence, if a child guesses too large a language, the world can never tell him that he is wrong." Consequence is that learning should be impossible. (NB: Gold proved this!). Obviously, though, children learn language, so there must be innate internal constraints on language learning that compensate for the lack of negative evidence. "Antoehr way of putting it is that it is puzzling that the English language doesn't allow "don't giggle me" and "she eated", given that children are tempted to grow up talking that way. If the world is not telling children to stop, something in their brains is, and we have to find out who or what is causing the change."

What Is Learned

What is language? Linguistics tells us this. One of its key discoveries is that there are language universals -- speicifically some structural properties of language that are independent of human experience. "For example, if a language has both derivational suffixes (which create new words from old ones, like -ism) and inflectional suffixes (which modify a word to fit its role in the sentence, like plural -s), then the derivational suffixes are alsys closer to the word stem than the inflectional ones are."

The role of universal grammar is to place constraints on the kinds of hypotheses about language that the learning mechanism will produce.

Input

What aspects of language do learners have access to? Grammatical sentences are sentences that sound natural in colloquial speech. Such sentences -- positive evidence -- comprise almost all of the language environment for children. "Thus, language acquisition is ordinarily driven by a grammatical sample of the target language."

"Negative evidence refers to information about which strings of words are not grammatical sentences in the language, such as corrections or other forms of feedback from a parent that tell the child that one of his or her utterances is ungrammatical. Behaviorist theories of language learning depend upon negative evidence. However, a variety of studies indicate that negative evidence is not provided in the linguistic environment.

Parental sppech to children is often simplified "baby talk" -- motherese. "One should not, though, consider this ... to be a set of language lessons. Though mothers' speech may seem simple at first glance, in many ways it is not." From a linguistic perspective, motherese still bombards children with complex structures. Also, evidence suggests that motherese does not affect rate of language acquisition.

Melody, timing, stress in speech -- called prosody. Prosody is correlated with syntax, and may be a source of information about syntactic structure during learning.

"Children do not hear sentences in isolation, but in a context. ... Many models of langauge acquisition assume taht the input to the child consists of a sentence and a representaton of the meaning of that sentence, inferred from context and form the child's knowledge of the meanings of the words." This can't be literally true, but some context can provide some meaning, and this is likely to help language acquisition.

What And When Children Learn

Child's language will be very similar to the language of their parents. How do we know taht children have mastered the intricacies of language? By performing clever experiments that reveal that complex rules underlie child's language. "Young children have grasped the abstract structural relations in sentences and have acquired a grammar of the same design as that spoken by their parents."

The Child's Language-Learning Algorithm

What algorithm do children use to induce a complex grammar from a language environment? "Innate knowledge of grammar itself is not sufficient. It does no good for the child to have written down in his brain `There exist nouns'; children need some way of finding them in parets' speech so that they can determine, among other things, whether the ouns come before the verb, as in English, or after, as in Irish." This is called the bootstrapping problem.

One solution to the bootstrapping problem would be to extract correlational data from language input. But there are serious problems with this view. First, it requires that the grammar is already learned, because the grammar would have to be used to identify features to correlate -- this defeats the whole purpose. Second, "without prior constraints ont he design of the feature-correlator, there are an astronomical number of possible itnercorrelations among linguistic properties for the child to test."

A second approach to solving the bootstrapping problem is to use prosody to parse input into syntactic components. This is problematic too, though, because prosody is affected by nonsyntactic factors, and therefore is not a reliable source of information about syntax.

A third approach to the bootstrapping problem "exploits the fact that there is a one-way contingency between semantics and syntax in the basic sentences of most of the world's languages. Though not all nouns are physical objects, all physical objects are named by nouns. ... If children assume that sematnic and syntactic categories are related in restricted ways in the early input, they could use sematic properties of words and phrases as evidence that they belong to certain syntactic categories." All that is required of this approach is that an initial set of rules is learned, because "once an initial set of ruels is learned, items that are more abstract or that do not follow the usual patterns relating syntax and sematnics could be learned through their distribution in already learned structures."

Acquisition In Action

"The child must couch rules in grammatical categories like noun, verb, and auxiliary, not in actual words. ... but words are not enough, they must be ordered." I.e., into phrases that have constituent structure. The search for phrases requires internal constraints.

"The use of part-of-speech categories, phrase structure, and meaning guessed from context are powerful tools that can help the child in the daunting task of learning grammar quickly and without systematic parental feedback. ...Indeed, if children are constrained to look for only a small number of phrase types, they automatically gain the ability to produce an infinite number of sentences, one of the hallmarks of human language."

"A grammar is not a bag of rules; there are principles that link the various parts together into a functioning whole. The child can use such principles of universal grammar to allow one bit of knowledge about langauge to affect another. E.g. the blocking principle that Pinker details in Chapter 5 of this volume.

Another example -- "every verb has an `argument structure': a specification of what kinds of phrases it can appear with. ... A general principle of argument structure is that the argument that is affected in some way (the particular way is specified by the verb) gets mapped onto the syntactic object." I.e., this is a linking rule between syntax and semantics. Mistaken interpretation of meaning (ie., argument structure of verb) leads to misapplication of the linking rule, that is revealed in very specific errors in children's language.

Striking discovery is that all languages are built on a common plan, and differ only in terms of setting choices for a small number of parameters. (NB: Creole is often described as a language that is the universal grammar with the set of default parameter settings.) "On this view the child would set parameters ont eh basis of a few examples from the parental input, and the full complexity of a language will ensue when those parameterized rules interact with one another and with universal parameters."

How do children learn parameters? "One suggestion is that parameter settins are ordered and that children assume a particular setting as the default case, moving to toher settins as the input evidence forces them to." This approach lets us get rid of the need for negative evidence. For instance, parameter ordering could be used to instantiate the subset principle:>/B> "children's first guess from among a set of nested possible langauges really is the smallest subset." Evidence is consistent with this principle.

Conclusion