Senin, 27 Desember 2010

The Acquisition of Syntax

The Acquisition of Syntax
Do children acquire language rapidly, or slowly? From the vantage point of linguistic theory, all normal children could be expected to have full command of a rich and intricate system of linguistic principles in just a few years. Experimental studies of child language, however, paint a different picture of language development: It appears that language learning extends over many years, with children making numerous missteps along the way. Attempts have been made to reconcile theory and data, by looking for features of language development that might impede the emergence of linguistic knowledge, and by looking in more detail at the experimental findings themselves. We begin by reviewing the reasons, based on current linguistic theory, for anticipating the rapid growth of linguistic knowledge in children. Then we turn to the laboratory, to consider both findings that do not sit well with the expectations of linguistic theory, as well as ones that comport well with theory.
a.      Why language acquisition is a snap?
Despite the complexity of human languages, children converge on a grammar that is equivalent to other members of their linguistic community; thereafter they are effectively adults in linguistic competence, including the ability to understand novel sentences, to discern relations of paraphrase and entailment, to judge that certain strings of words are ill-formed, and to judge that certain well-formed strings of words lack meanings that they might be expected to have, based on related examples. This remarkable acquisition scenario unfolds every day, all across the globe, yielding the truism that any human child can learn any human language.
The road-map children use to plot a course in grammar formation enables them to project beyond their experience, rather than being securely tied to it. Children are therefore expected to form grammars that deviate in certain respects from those of adult speakers of the target language. But, like Rome, all roads lead to the same destination; at some point, children achieve a stable state that is equivalent to that of adults in the linguistic community. From this perspective, the errors that arise in the course of language acquisition are not the result of defective grammars rather, language-learners sometimes speak a foreign language (metaphorically speaking).  This is the continuity assumption, the proposal that child language can differ from the language of the local community only in ways that adult languages can differ from each other. As a general research strategy, advocates of the continuity assumption suppose that explanations of differences in behavior between 2 children and adults should invoke minimal differences in cognitive mechanisms, including linguistic principles.
Experience matters on the kind of (Principles and Parameters) model we are assuming. After all, children exposed to English learn English, and children exposed to Basque learn Basque. Once children achieve a stable state, however, there appear to be no lingering reminders of any wrong turns that children might have taken in the course of grammar formation. It is as if each child had access to all the primary linguistic data at once, and had reached the stable state instantaneously. The idea of 'instantaneous acquisition' is an idealization, of course. But the fact that the actual course of language development does not leave any indelible marks on the adult grammar is important to linguistic research. This fact permits linguists to investigate the grammars of human languages without concern for any childhood "errors" that are made in language development.
Any missteps in setting parameters must be set straight, to redirect children to a grammar that is equivalent to that of adults, and setting a new course requires (positive evidence) from other speakers of the local language. So, depending on the properties of the input available to children, it is conceivable that children spend some amount of time using grammars that differ in certain respects from those of other speakers, and more like grammars used by speakers of other human languages. But how long should such 'stages' last before children find their way back on track? If parameter values can be revised on the basis of simple and readily available features of every child's linguistic experience, then the stages of language development should not last long at all. And the logic of the situation dictates that the available evidence for parameter setting must be both simple and abundant. If the evidence needed for parameter setting were exotic or required excessive computational resources, then some (perhaps many) children would not encounter the requisite data, or would be unable to 3 perform the necessary computation. These children would not successfully converge on an adult grammar. Of course this does not happen: all normal children learn to speak the language of the local community. Therefore it is safe to conclude that the evidence for parameter resetting is simple and readily available. So, ceteris paribus, children should rapidly converge on the target grammar.
It should be clear, now, why rapid acquisition is anticipated by recent developments in linguistic theory in the Principles and Parameters framework, along with certain assumptions about the nature of the input. The findings from some experimental studies of child language, however, suggest that children stabilize on an adult grammar only after several months, even years. In the next section, we examine one finding that has been interpreted as demonstrating the late emergence of linguistic knowledge. To complete the picture, however, later sections report findings that are consistent with the rapid acquisition scenario that is expected according to linguistic theory.
b.      The Emergence of Relative Clauses
An example of the apparently late acquisition of syntactic knowledge was uncovered in research conducted in the 1970's. In several studies children were found to commit systematic errors in responding to sentences with a restrictive relative clause. Children's errors appeared in experiments using an 'act out' methodology. These studies found that four and five-year-old children consistently acted out sentences like in a non-adult fashion for example:
1.      The dog pushed the sheep that jumped over the pig.
2.      The dog pushed the sheep and jumped over the pig.
When asked to act-out the meaning of the first sentence, the majority of the child subjects had the dog push the sheep and then jump over the pig. For adults, it is understood to mean that the sheep jumped over the pig. Children's non-adult responses led researchers to claim that children assigned a structure that was appropriate for conjoined clauses, as if sentence first sentence had the structure appropriate for second sentence, according to which the dog, not the sheep, jumped over the pig. Accordingly, this proposal was called the Conjoined Clause Analysis. For second sentence, using other experimental procedures, it was demonstrated that English-speaking children and Italian-speaking children have mastery of sentences with a relative clause at a much younger age, even before their third birthday The innovation in procedure was motivated by the observation that a sentence with a restrictive relative clause, such as first sentence, bears two kinds of presuppositions. First, felicitous use of the NP, the sheep that jumped over the pig, presupposes that there are at least two sheep in the conversational context. If there is only a single sheep, there is no need to add a modifier; the speaker could just as well have said "The dog pushed the sheep." In short, a restrictive relative clause is appropriate when some restricting needs to be done. The second presupposition of the first sentence is that the event mentioned in the relative clause (the jumping event) should have taken place prior to the event mentioned in the main clause (the pushing event).
c.       The Head Movement Constraint
The next example illustrates the claim that children do not venture beyond the boundaries established by Universal Grammar. Indeed, the space of possible human languages is so restricted as to exclude even apparently simple rules, and ones that are compatible with much of the primary linguistic data available to children. For examples:
a) Bill can play the sax.            Can Bill play the sax?
b)  The sky is blue.                   Is the sky blue?
However, the simple structure-independent hypothesis move the first {is, can, will, …}, produces the wrong Yes/No questions for sentences that contain a relative clause, such as example number 4. In the Yes/No question corresponding, to number 4, the auxiliary verb, is, following the entire subject NP (noun phrase), the man who is feeding a donkey, has to move, as in example number 5. Moving the first occurrence of is, from inside the relative clause (who is feeding a donkey), results in a deviant Yes/No question, example number 6.
(4) The man who is feeding a donkey is mean
(5) Is the man who is feeding a donkey _ mean?
(6) Is the man who _ feeding a donkey is mean?
d.  Why-questions: A Case Study in Continuity
As mentioned in the introduction, the continuity assumption supposes that children and adults share a common core of linguistic knowledge. Essentially, child language can differ from adult language only in ways that adult languages can differ from each other. So far, the literature that we have reviewed on child language is consistent with the continuity assumption: children do not appear to entertain grammatical hypotheses that extend beyond the boundary conditions imposed by Universal Grammar. Advocates of a Principles and Parameters theory should not be surprised, however, if some English-speaking children exhibit some features of grammar that appear in other Germanic languages, Romance or East Asian languages, in the absence of evidence for these properties in the primary linguistic data. Evidence of children’s non-adult (but UG-compatible) productions may be the strongest argument for the theory of Universal Grammar. An argument of this kind can be made using evidence from the acquisition of Why-questions. Examples:
Why you have your vest on?                Why she doesn’t have any hanger?
 Why he’s woofing?                             Why that guy has tookened Walker?
Why he’s following the guy?               Why that kind of thing could break?.
e.       Avoiding errors: innate constraints versus conservatism 
A constraint on reference. Another distinguishing feature of the two approaches to language development is how they explain the kinds of sentences children refrain from producing, and the kinds of meanings that children do not assign to sentences. One case in point is the reference of ordinary pronouns. Notice that in the examples number 7 and number 8, the pronoun he may or may not refer to the individual called the Ninja Turtle. To indicate these dual referential possibilities, we will adopt the following notation: two expressions refer to the same individual only if they have the same index. So, in examples number 9 and 10 are ambiguous, because the pronoun he can have the same index as the Ninja Turtle (1), but one of these expressions can also be assigned an index (2) which the other expression lacks in that case, the two expression are said to be disjoint in reference or non-coreferential.
7.   The Ninja Turtle1 danced while he1/2 ate pizza.
8. While he1 ate pizza, the Ninja Turtle 1/2 danced. 
Consider another sentence number 9, which also contains the pronoun he and the expression the Ninja Turtle. Unlike examples number 7 and 8, 9 is unambiguous. Intuitively, the pronoun he cannot refer to the Ninja Turtle, but must refer to some other male individual. In other words, co-reference (as indicated by the assignment of the same number) is not permissible in the sentence in 9, it has the reading in 9a, but not the reading indicated in 9b. Examples:
  9. He danced while the Ninja Turtle ate pizza. 
(a)   He1 danced while the Ninja Turtle2 ate pizza 
(b) *He1 danced while the Ninja Turtle1 ate pizza 

Children's optional infinitive errors are a potential problem for the continuity assumption, because root infinitives are not a core phenomenon of adult languages. On several current accounts, optional infinitives arise because the young child’s grammar has not matured, such that children can optionally omit tense or agreement features from a phrase structure representation (e.g., Schütze and Wexler 1996). Omission of agreement results in an utterance with default (accusative) case; omission of tense results in children’s failures to produce present or past tense marking. Another maturational account attempts to further minimize the differences between child and adult grammars. The only difference between child and adult grammars, on this account, is that children do not require clauses to project up to the CP level at this stage of development; until this knowledge matures, children allow ‘truncated' structures (Rizzi 1993). A third account is more in keeping with the continuity assumption). On this account, children’s syntax is intact. Optional infinitives are the product of a processing bottleneck that sometimes prevents children from merging the inflection and the main verb (Phillips 1995. Only time will tell whether the optional infinitive phenomenon represents a genuine violation of continuity, or if it can be squared with the continuity assumption.
            Empirical investigations in the acquisition of syntactic knowledge have reached a new level of maturity, in just the past ten years.  If developments in methodology and theory keep pace, we will continue to achieve a deeper understanding of children's universal mastery of syntactic competence, and we will find out if the continuity assumption can be maintained.



The example is Chomky's (1971; 1975) parade case of 'the logical problem of language acquisition,' the formation of Yes/No questions



















References
Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, MA: The MIT Press. 
M. Traxler and M. Gernsbacher (eds), Handbook of Psycholinguistics, Elsevier.
http//www. 2ECS.suntax 2010.com

0 komentar:

Posting Komentar