4.2 Current approaches to the study of language acquisition
4.2.1.2 Content of the Language Faculty
Scientific study of the Language Faculty requires specification of its content. This has developed with the science of linguistics.
4.2.1.2.1 Chomsky’s LAD
Chomsky’s early formulation of a Language Acquisition Device (LAD) logically explicated the preconditions for acquiring linguistic knowl- edge on the basis of projection from input (Chomsky 1984, 30; Chomsky 1999, 43), thus beginning to formalize a solution to the Projection Problem raised in chapters 2 and 3 (Peters 1972, 173).
4.2.1.2.2 Challenges to LAD
Formulation of the LAD appeared to beg the issue of language acqui- sition. Component 3 was often interpreted as implying that a predetermined set of “specific language grammars” were innate (English, Swahili, Sinhala, Hindi, etc.) and that these merely needed to be “selected from,” raising the question of how these grammars arise and how children judged whether the data were “com- patible” with the grammar hypothesized (exemplifying the “evaluation metric”
(component 5)) (Peters 1972, 179). The model in table 4.3 is a-temporal, consistent with an “instantaneous” view of language acquisition an assumption Chomsky realized is “obviously false” (1975, 119, 121).
4.2.1.2.3 From LAD to UG
Chomsky moved the theory of the Language Faculty from LAD to Universal Grammar (UG).
4. “[U]niversal grammar is part of the genotype specifying one aspect of the initial state of the human mind and brain.” (Chomsky 1980, 82)
6See Chomsky 1975, 122 on interpretation of the innateness hypothesis.
54 c h i l d l a n g ua g e
Table 4.3 Chomsky’s early model: Language Acquisition Device (LAD) (Aspects of the Theory of Syntax, 1965; chapter 1, 30–31).
Components of LAD These components require
1. a technique for representing input signals
1. a universal phonetic theory that defines the notion “possible sentence”
2. a way of representing structural information about these signals
2. a definition of “structural description”
3. some initial delimitation of a class of possible hypotheses about language structure
3. a definition of “generative grammar”
4. a method for determining what each such hypothesis implies with respect to each sentence
4. a method for determining the structural description of a sentence, given a grammar
5. a method for selecting one of the (presumably, infinitely many) hypotheses that are allowed by 3, and are compatible with the given primary linguistic data
5. a way of evaluating alternative proposed grammars
The definition of UG is formal, general and abstract, no longer suggesting access to a list of pre-defined grammars. UG differs from LAD in its formulation of what is proposed to be biologically programmed. Consequently, it reassesses the relation between children and the PLD.
5. “Universal grammar may be thought of as some system of principles, common to the species and available to each individual prior to experience”
(Chomsky 1981b, 7).
UG is bidimensional, as in (6).
6. Universal Grammar
a. “Universal Grammar might be defined as the study of the conditions that must be met by the grammars of all human languages” (Chomsky 1968, 126, 62)
b. “In a highly idealized picture of language acquisition, UG is taken to be a characterization of children’s pre-linguistic initial state” (Chomsky 1981, 7) and of the “language faculty” (Chomsky 1981, 7).
If all natural languages follow a universal architecture, and the human species is so programmed, this would explain why universals of language exist and why any language is normally acquired in children’s first years, escaping the episte- mological dilemma seen in chapters 1–3.
4.2.1.2.4 Armed for discovery
Theoretical developments over the last several decades have progres- sively: (a) streamlined the theory of Universal Grammar;7 (b) eliminated many
“substantive” proposals and replaced them with more abstract “formal” univer- sals, substituting general principles for specific “rules”;8 (c) shifted focus away from the question of how linguists or children analyze a sentence with a specific structure, and toward a focus on identifying and understanding the underlying architectural principles which converge to result in knowledge of specific con- structions.
“Principles and Parameters” theory of UG defines both a set of universalprin- cipleswhich capture what underlies language structure everywhere, and a finite set ofparametersto account for possible cross-linguistic variation.
4.2.1.2.4.1 Principles The most fundamental principle of UG is “structure dependence.”9Armed with it, children acquiring a language will not be thwarted by the continuous speech stream, but will impose linguistic units upon it. Children will never consider language data as simply linear “beads on a string,” but will know that grammatical computation depends on structure.
7. Structure dependence
“The rules operate on expressions that are assigned a certain structure in terms of a hierarchy of phrases of various types” (Chomsky 1988a, 45).
This principle provides children with the foundation for discovering and building the secret skeleton that underlies every sentence. Neither will children be thwarted by the fact that linguistic elements move or disappear, since they now have the basis for tracking movement of units and knowing where they might or might not appear.
The principle of structure dependence rules out infinite possible false hypothe- ses regarding possible computations. In a classic example, Chomsky argues that without structure dependence and by induction (“analogy”) alone on sentences like (8) in English or (9) in Spanish, a child might surmise a rule such as: delete the first “is” or “est´a” in the sentence; insert one of these at the beginning of the sentence; thus leading from examples like (8a) or (9a) and (8b) and (9b) to the ungrammatical (8c) or (9c) (Chomsky 1988a, 41f).
8. English
a. The man is in the house Is the man in the house?
7The acronyms TG (Transformational Grammar), GB (Government and Binding Theory) and P&P (Principles and Parameters Theory) reflect developing versions of Universal Grammar.
8The distinction between “substantive” universals and “formal” universals is difficult. The basic distinction intended is: universals that specify specific content, e.g., nouns or verbs vs. universals that refer to the architecture by which symbolic forms are generated and computed over.
9UG also would provide principles and parameters for the areas of semantics and phonology, although these areas have received less work in theory and in language acquisition.
56 c h i l d l a n g ua g e b. The man is happy
Is the man happy?
c. The man who is in the house is happy
* Is the man who in the house is happy?
9. Spanish
a. Est´a el hombre en la casa?
Is the man in the house?
b. Est´a el hombre contento?
Is the man happy?
c. *Est´a el hombre que contento est´a en la casa?
Is the man who happy is in the house?
Structure dependence constrains children’s grammatical hypotheses so that they must analyze the structure (bracketed) in (10) and (11) before forming a question inversion rule and thus not first “consider the simple linear rule . . . then discard it” (Chomsky 1988a, 45). The structure-dependent rule comes first. Children’s knowledge would be to a degree deductive; knowledge acquisition would be initially and continuously constrained.
10. [The man[who is in the house]] is happy [El hombre [que est´a contento]] est´a en la casa 11. Is [the man [who is happy]] at home?
Est´a [el hombre [que esta contento]] en la casa?
4.2.1.2.4.2 Parameters Parameters target critical dimensions for grammar building and constrain cross-linguistic variation along these dimensions.
12.PARAMETER: a principled dimension of language variation, which specifies predetermined values of this variation. Parameters provide the “atoms” of linguistic structure (Baker 2001).
In the strongest theory, parameters are minimal in number and binary valued.
When set, parameters allow children to establish the basic forms of grammars and draw widespread deductive conclusions from this.
13. UG Parameters: A Switch-box Metaphor
“The initial state of the language faculty consists of a collection of
subsystems, or modules as they are called, each of which is based on certain very general principles. Each of these principles admits of a certain very limited possibility of variation. We may think of the system as a complex network, associated with a switch box that contains a finite number of switches. The network is invariant, but each switch can be in one of two positions, on or off. Unless the switches are set, nothing happens. But when
the switches are set in one of the permissible ways, the system functions, yielding the entire infinite array of interpretation for linguistic expressions.
A slight change in switch settings can yield complex and varied
phenomenal consequences as its effects filter through the network . . . To acquire a language, children’s mind must determine how the switches are set” (Chomsky 1988, 68).
Example parameters Word order: head direction
One language may demonstrate word (and constituent) orders that reverse another language (cf. chapter 2). Through a “head direction” parameter, UG allows languages to vary order (chapter 9). This setting will generalize across various constituents. To a significant degree, children can then deduce basic aspects of their language.
Pro Drop
Romance languages, among many others, allow subject omission as in (14) (Haegeman 1991, Chomsky 1986) and are categorized “+Pro Drop,” unlike English (15). Parameter setting of a language as (+) or (–) “Pro Drop” would direct children to this dimension of variation and allow cross-linguistic varia- tion in a principled manner. If children set the Pro Drop parameter, they can, to some degree, determine grammatical consequences by deduction. For example, expletive “it” subjects, as in “it is raining”, may occur in non-Pro Drop languages.
14. a. Ø ha telefonato
b. Giacomo ha detto che Ø ha telefonato 15. a. John has telephoned
* Ø has telephoned
b. John has said that he has telephoned
* John has said that Ø has telephoned
4.2.1.2.5 Summary
UG provides the basis for a more comprehensive theory than the clas- sic formulation of the LAD, i.e., one which relates linguistic theory to language development in real time.
a. Recognizing the necessity for biological programming of initial lin- guistic knowledge, UG does not propose that specific language gram- mars are innate, only the universal architecture for language.
b. UG does not deny that other cognitive components are necessary to language acquisition. It is one “specific component of LT(H, L)(Learn- ing Theory of a human for language) (Chomsky 1975b, 28).
c. UG does not deny the role of experience. It proposes that abstract linguistic principles interact with experience and allow children to go beyond the limits of actual experience.
58 c h i l d l a n g ua g e
d. UG eliminates an “evaluation metric” of the form in the original LAD (Chomsky 1996, 171). Principles and parameters direct and constrain children’s experience so that their interaction with the PLD can deter- mine a language.10
e. UG allows us to bridge deduction and induction throughabduction.11
“Abduction” refers to the analysis of observed fact against explanatory hypotheses. A model including abduction recognizes the necessity for children to use the input with which they are presented, but allows that they do so in a guided way (by the Language Faculty), going beyond the limits of induction alone. Peirce exemplifies this form of inference in (16) (1955, 151).
16. Abduction
i. A surprising fact, C, is observed.
ii. But if A were true, then C would be a matter of course.
iii. Hence, there is reason to suspect that A is true.
4.2.2 Challenges to the UG paradigm
Given the armamentarium of a Language Faculty, we must still ask:
(a) what is the relation of children to the input data and (b) what is the nature of development? How are UG “principles” applied, and “parameters” set? What are the “primitives” of the Language Faculty, and how are they realistically applicable to children’s input data?
How can children acquire a grammar or set a parameter without having a grammar by which to process the input data? If children are to use input, they must be able to parse it. Perhaps “parser failure” is informative, e.g., if the parser fails, it tells children that the data are not compatible with their hypothesis. Here children are assumed to be “error-driven.” However, are there not “inherent limitations on children’s parser” (Valian 1990, 144)? If parsing is deficited, how can children attain useful data? If parsing is determined by a grammar for a specific language, then how can children begin the process of using input data or deal with data from another language in order to set a parameter?12