Child Language Acquisition Essay Example

The goal is to analyze the Universal Grammar (UG)-specific constraints on knowledge, and their ability to aid language learners. They ultimately conclude that UG-specific constraints do not aid in language learning. 

UG is being defined here is what it means in the field of child language acquisition. Ambridge et al. defines UG  as a set of constraints of language that are genetically encoded or innate.

The constraints of UG that they identify are:

Identifying Syntactic Categories

Acquiring Basic Morphosyntax

Structure Dependence

Subjacency

Binding Principles 

For each component, the authors identify several theories from UG about how children acquire these components; then establish which problem(s) these theories suffer from.

They identify 3 problems that the different constraints on innate knowledge suffer from.

LINKING

INADEQUATE DATA COVERAGE

REDUNDANCY

Based on these problems, the authors conclude that there are no proposals, currently, for a type of innate knowledge that aids in language learning.

The authors state their reasoning for this paper as the fact that UG is an active hypothesis that many researchers in both linguistics and child acquisition incorporated into their work. The authors do not take issue with the hypothesis of innate knowledge, but rather they believe that the components discussed below are common-held beliefs that have not been thoroughly examined in the context of actual language learning.

Their goal is not about drawing conclusions, but rather to challenge advocates of UG to have better data coverage and to quantify what the theory of UG is. 

The two components of particular importance that will be discussed in detail are structure dependence and binding principles.

Structure Dependence

Structure Dependence in UG  is defined as the knowledge that rules must reference syntactic structure, not linear order. 

Structure dependence is seen as a component of UG because young children are able to produce correctly formed questions, which would indicate to UG advocates that they have innate knowledge about language generally being structure dependent. The article discusses structure dependence in the context of complex  yes-no questions, as well as a general overview of structure dependence.

4.1   Complex Yes-No Questions

Suffers from inadequate data coverage.

There are two questions Ambridge et al. examined in this section.

How do children acquire the correct generalization and avoid errors?

How do children know that all linguistics generalizations are structure-dependent?

In response to these questions, 3 possible answers are presented, answers that do not require innate knowledge.

The first answer  is that questions are not formed by movement rules. 

The second answer  is that language learners are sensitive to the fact you cannot remove parts of an utterance that contain background information. Evidence for this is given in (6) and (7).

(6) a. Speaker 1:  The boy [who is smoking] is crazy.

        b. Speaker 2: No, sane. *No, drinking beer.

(7) Is the boy {who is smoking]__ crazy? Vs. *Is the boy {who__ smoking] is crazy?

7 demonstrates how only parts of the main clause can be removed and remain grammatical.

The final answer only applies to complex yes-no questions in English and states that children use of bi/trigram statistics.  Evidence for this solution is given from a study done by two other researchers, Reali and Christiansen; as well as a  computer simulation to predict the correct order. The third answer is overall the weakest and Ambridge, Pine and Lieven note that the model of bigram statistics only works because of how Engish complementizers and WH-words/deictic pronouns work. This does not work cross linguistically.  

4.2  Structure Dependence in General

After reviewing the case of complex yes-no questions, structure dependence in general is discussed. The article opens with an example to provide a context for the discussion of how children learn syntactic rules are structure dependent. 

The example is:

(8) a. John is smiling. Yes, he is happy.

        b. The (/that/this/a etc.) boy is smiling. Yes, he is happy.

        c. The tall boy is smiling. Yes, he is happy.

        d. The boy who is tall is smiling. Yes, he is happy.

Ambridge et. al argues that children are exposed to phrases like (8) thousands of times a day and uses these examples as evidence for how a learner acquires a structure-dependent grammar. The sentences in (8) are evidence for the claim that strings of any length that share similarities in distribution, can be exchanged with each other. John, the tall boy, he can all be exchanged for each other because they all refer to entities that are semantically related in content. Consequently, they can undergo the same semantic operations. Ambridge et al. relates this concept to the fact that languages group together physical nouns, like John and abstract nouns like idea is  because they can undergo the same functional operations. Therefore in order to acquire a structure-determined grammar a learner has to understand that words like John, the tall boy and idea share functional similarities. This understanding, when applied to a language will yield a structure-dependent grammar, according to Ambridge et al.

Next a study done by Crain and Nakayama(1987) is discussed. The study found that children showed identical performance for questions with semantically empty subjects.

This appears to be in opposition to the claim made by Ambridge et al; however, Ambridge et al provide three options for what the children could have learned already as evidence for their claim.

Learned the formulaic  questions, such as Is there a(n) [THING]? Is it easy to [ACTION]?

Learned that Is it and Is there are common ways to begin a question.

Generalized between dummy and lexical subjects on the basis of functional and distributional overlap.

To end their discussion of structure dependence, the authors address the subject of syntactic category labels, and how these labels are not inherently UG. Instead, stating that perhaps it is the functional categories (nouns, verbs etc.) do not work cross linguistically. The authors purpose that some biases; such as those mentioned in the Crain and Nakayama study, may be assumed. But ultimately they conclude this idea of an innate principle is different from the idea of innate structure dependence.    

6. Binding Principles

Another key component discussed by Ambridge et al is binding principles. Languages block certain pronouns from referring to certain NP’s. An example of this is given in (24).

(24) She listens to music when Sarah reads poetry.

She cannot refer to Sarah.

The UG explanation of this is that this knowledge is unlearnable and must be specified by innate “binding” principle. Binding is defined as X binds to Y if X c-commands Y AND X & Y refer to the same entity. There are three binding principles, A, B and C. the Ambridge et al begin with Principle C.

The authors intend to showcase that Principles A, B and C suffers from the problem of redundancy as they claim all the principle can be explained through other principles.

6.1   Principle C

Principle C states that (for multiple-clause phrase at least) a pronoun might occur before a full lexical NP to which it corefers. Only if the pronoun is in a subordinate clause.

This article proposes that Principle C only works up to the extent that it corresponds with principles of information and discourse. A functional explanation is given as the theme is the NP that the sentence is making an assertion about. The assertion is made in the predicate of the main clause, such as Sarah listens to music. When what the referent is referring to is known (known as topic) it is most natural for speakers to use a pronoun, as in Sarah listens to music when she studies. Based on this, when speakers use lexical NP as topic, they intend to establish this NP as the topic of a new referent. Once the new referent has been established it is more natural for speakers to use a pronoun to refer to it.

This generalization about the constraints Principle C places on languages, accounts for all the data except for the case of forward anaphoras from a subordinate into a main clause. An example of this is when Sarah reads poetry, she listens to music.

This exception cannot be explained by syntactic principles but can however be explained using the  discourse-functional principle proposed by Ambridge, Pine and Lieven. They use the following examples as evidence for this claim.

(31) a. Sarah reads poetry. She also listens to music.

        b. *She reads poetry. Sarah also listens to music.

Ambridge et al account for the data in (31) by stating that no binding principles can apply between sentences or conjoined clauses. Therefore only the discourse-functional principle outlined above can account for the pronominalization that occurs between these sentences.

Finally Ambridge, Pine and Lieven conclude that children need to know certain discourse-functional principles and that the discourse-functional principles they outlined account for all the data. This explanation of the data makes syntactic UG-based accounts redundant. 

Principles A and B

Principle A states that a reflexive pronoun must be bound in its local domain which is its clause. It can only refer to an NP that c-commands it in the clause.

Principle B is the opposite of Principle A. It states that a non reflexive pronoun must be free.

It is argued that binding principles A and B can be explained using the functional principle mentioned previously. The functional principle explanation is reflexive pronouns are used in English only if they are direct targets or recipients of the actions demonstrated in the phrase. 

The following is three sentences from (33) that is provided as evidence:

 (33) a. John [killed/fell in love] with himself/*him. (target)

         b. John addressed the letter to himself/*him. (recipient)

         c. John [heard strange noises/left his family] behind *himself/him. (location)

UG principles A and B are redundant, according to Ambridge et al not only because they make very similar assumptions, but also because they fail to account for data such as in (35).

(35) Q: Who did Sue say is the cleverest girl in the room

        A: Herself(*Her)

The functional principle can apply across sentences and thereby eliminate the impossible reading. 

Ambridge et al also point out that principle B is violated in ‘Evans-style’ contexts, where, in the example (38), him is c-commanded by he. The functional principle explanation for this violation is that him is used because the meaning of the sentence is ‘the person who may be John looks like John’ therefore he looks like him is acceptable.

This argument is closed by reinstating that because discourse functional principles are needed to explain counterexamples, syntactic principles are rendered redundant.  

Conclusion

The authors have shown that all the components of UG  suffer from linking, data coverage or especially redundancy.

The primary claim is that there is no current UG proposal that demonstrates that innate knowledge aids language learning.

The authors identify possible criticism by stating that even though they did not propose an alternative to UG, it is still is not logical to continue using UG. 

Instead they propose UG advocates focus more on applying their theories by indicating what can be learned with UG and more importantly, how a child could do so.

I do think  that an alternative to UG should have been presented especially because the claim being established is very bold. The authors reinstate that their goal was only to show that UG-based theories do not aid language learners; but this reinstatement of their argument upon criticism does make it weaker. A truly strong argument against UG would have included a framework for an alternative hypothesis; one that they could present in the face of such criticism and one that is more than simply a challenge to the opposition.

Despite the lack of conclusions drawn in this article, I do think the article discusses a very important topic; that being essentially does UG really work for child development? It is rather weak in that it does not attempt to provide answers but it is a decent starting place for developing a more comprehensive version of UG.