A Concise Introduction to Syntactic Theory

The Government-Binding Approach

Elizabeth A. Cowper

• Publisher: University of Chicago Press
• Year: 2009
• Print ISBN: 9780226116464, 9780226116440
• eBook ISBN: 9780226160221
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This textbook is intended to give students a quick start in using theory to address syntactic questions. At each stage, Cowper is careful to introduce a theoretical apparatus that is no more complex than is required to deal with the phenomenon under consideration. Comprehensive and up-to-date, this accessible volume will also provide an excellent refresher for linguists returning to the study of Government-Binding theory.

"Cowper exhibits the analytical devices of current principles-and-parameters approaches, takes readers carefully through the central elements of grammatical theory (including very recent work), and ushers them selectively into the technical literature. . . . A serious introduction for those who want to know the nuts and bolts of syntactic theory and to see why linguists are so excited these days."—David Lightfoot, University of Maryland

"An excellent short introduction to the Government and Binding model of syntactic theory. . . . Cowper's work succeeds in teaching syntactic argumentation and in showing the conceptual reasons behind specific proposals in modern syntactic theory."—Jaklin Kornfilt, Syracuse University

1 The Theory in Context

The theory of Government and Binding has developed out of a tradition going back to the 1950s, known variously as transformational grammar (TG), generative grammar, or generative-transformational grammar. While it is not the purpose of this textbook to give a detailed, or even thorough, history of the theory, I will give a short outline of how the theory has changed and developed over the years, so as to provide a context for what follows.

In section 1.1, some of the most fundamental goals and assumptions of the theory of generative grammar will be identified. These are the things which have remained constant throughout the history of the theory, even though just about every aspect of how the theory is structured and implemented has changed. In section 1.2, a number of major stages of the theory will be identified, and the major properties of each stage will be described. For a more comprehensive discussion of the development of the theory, see Newmeyer (1980).

1.1 Goals and Assumptions

The fundamental problem of linguistic theory, according to Chomsky, is that of “determining how it is possible for a child to acquire knowledge of a language” (Chomsky 1973:12). This question has remained at the core of work in generative grammar since its inception. In order to answer this question, we must first examine it and make sure we understand it precisely enough that it can guide our investigations in a meaningful way.

First, what is meant by “knowledge of a language”? I am not speaking here of the kind of explicit, conscious knowledge taught in elementary-school grammar classes. Rather, I mean the largely unconscious knowledge that makes us speakers of a language—the knowledge we use when we judge that (1) is a grammatical sentence of English while (2) is not.

(1)	Mary is dancing on the stage.
(2)	*Mary are danced the stage on.

Before we can begin to answer the question of how knowledge of a language can be acquired, we must have some notion of exactly what it is that is being acquired.

Knowledge, however, is not something we can observe directly. This is especially true in the case of language. Every normal human being is a native speaker of (at least) one language and thus by definition has acquired knowledge of that language. Most people, however, never study their native language in any conscious way. Just as people know how to walk without consciously knowing which muscles, nerves, and parts of the brain are involved, people know their native language without consciously knowing its structure. In contrast, people who know predicate logic, or chess, for example, normally do have an awareness of the structure of the system of rules governing what can be done in logic or in a game of chess.

Since we cannot observe knowledge of language directly, how then can we study it? What we can do is observe people as they use this knowledge in various ways—as they speak and understand their native language. In other words, we can observe the linguistic behavior of native speakers. Another thing we can do is ask people, including ourselves, to use their knowledge in judging whether particular sentences are acceptable sentences of their native language. From these types of linguistic behavior, we can then try to deduce the knowledge that enables them to perform the behavior. S. Jay Keyser, in class lectures in the late 1960s, put it very well: We are trying to figure out what it is that people act as if they know. Our job is therefore not merely to describe what people say, but, rather, to figure out what might be the knowledge which permits them to perform their linguistic behavior. We shall henceforth refer to this knowledge as the speaker’s linguistic competence and to the behavior which we can observe as the speaker’s linguistic performance.

The problem, of course, is that linguistic competence is not the only factor which influences linguistic performance. For this reason, not everything a native speaker of English says is an equally reliable indicator of that speaker’s linguistic competence. A rather blatant example is given in (3).

(3) Please don’t shut the window on my [loud scream].

External events, such as a window shutting on someone’s hand, can interrupt a speaker and force the abandonment of a sentence in midstream. No one would seriously propose that the sentence in (3) as it stands constitutes a grammatical sentence of English. Rather it is a sentence fragment, or an incomplete sentence, which happened to be uttered by someone on a particular occasion.

While it is fairly clear that (3) can be discarded as contaminated data, many cases are far less obvious. Consider, for example the sentences in (4) and (5).

(4)	a.	They talked to Sue and I about the accident.
	b.	Me and Sue saw the accident.
(5)	a.	They talked to I about the accident.
	b.	Me saw the accident.

Sentences like those in (4) are produced by speakers of English fairly frequently, while sentences like those in (5) are almost never observed. Nonetheless, most speakers of English would say that all four of the sentences are ungrammatical. The problem with all of these sentences has to do with the form of the pronoun I/me. Normally, when this pronoun occurs as the object of a verb or of a preposition, it takes the so-called objective form, me. When it occurs as the subject of a clause, it normally takes the so-called nominative form, I. Confusion tends to arise when this pronoun occurs in a coordinate structure containing the conjunction and. The question is, in constructing a grammar of English which is supposed to reflect the competence of native speakers of English, do we consider the sentences in (4) grammatical or ungrammatical? If we consider them grammatical, then the rule governing the choice of pronoun form will have to have in it a special subclause saying that in coordinate structures, the choice of form is freer. If we consider them ungrammatical, then we must explain why speakers often produce sentences like (4) and almost never produce sentences like (5).

The point here is that in order to develop a theory of competence, or a model of a native speaker’s linguistic knowledge, we will, at every step of the way, be making judgments about the relevance of the data. These judgments are possible only in the context of the theory itself. The theory of linguistic competence will ultimately interact with other theories of memory, production, and comprehension, as well as with an understanding of external events (see (3) above) to account for particular instances of linguistic behavior. For the moment, we will simply say that generative grammar has been and continues to be primarily concerned with linguistic competence.

The next question that arises in our examination of Chomsky’s question has to do with the model of knowledge, or grammar, we are constructing. What should the grammar do? According to Chomsky, the grammar must explicitly account for all of the grammatical sentences of the language under consideration. In other words, every grammatical sentence of the language must conform to all the requirements of the grammar, and every ungrammatical sentence must violate some requirement of the grammar. In this, generative grammars differ from the traditional and structuralist grammars that preceded them. Those grammars, again according to Chomsky,

do not attempt to determine explicitly the sentences of a language or the structural descriptions of these sentences. Rather, such grammars describe elements and categories of various types, and provide examples and hints to enable the intelligent reader to determine the form and structure of sentences not actually presented in the grammar. Such grammars are written for the intelligent reader. To determine what they say about sentences one must have an intuitive grasp of certain principles of linguistic structure. These principles, which remain implicit and unexpressed, are presupposed in the construction and interpretation of such grammars. While perhaps perfectly adequate for their particular purposes, such grammars do not attempt to account for the ability of the intelligent reader to understand the grammar. The theory of generative grammar, in contrast, is concerned precisely to make explicit the contribution of the intelligent reader. (Chomsky 1973:8)

When a particular sentence conforms to all the requirements, or rules, of a generative grammar, we say that the grammar generates that sentence. If a sentence violates one or more requirements or rules of the grammar, then we say that the grammar fails to generate that sentence. If a grammar generates all the grammatical sentences of a language, and fails to generate any ungrammatical sentences, then we say that the grammar is observationally adequate—it successfully distinguishes between grammatical and ungrammatical sentences.

It is entirely possible that several very different observationally adequate grammars could be written for the same language. The goal of linguistic theory, however, goes beyond simply describing which sentences are grammatical and which are not. What we are trying to understand is not the language at all, but knowledge of language and how it can be acquired. Our grammar must therefore achieve more than observational adequacy. In addition to accounting for grammatical versus ungrammatical sentences, it must capture linguistically significant generalizations. In other words, it must provide the correct analysis for the sentences of the language—the analysis which corresponds to the native speaker’s (unconscious) knowledge. For example, if two sentences are related to each other in a systematic way, as are the sentences in (6), then the grammar must explicitly account for that relation.

(6)	a.	Sue started the car.
	b.	The car started.

A grammar which accurately reflects the native speaker’s knowledge is called a descriptively adequate ...

Table of contents

• Cover
• Copyright Page
• Title Page
• Dedication
• Contents
• Preface
• 1: The Theory in Context
• 2: Categories and Phrase Structure
• 3: Thematic Relations and θ-Roles
• 4: Predicting Phrase Structure
• 5: NP-Movement
• 6: Government and Case
• 7: WH-Movement
• 8: Move α and the Theory of Movement
• 9: The Empty Category Principle
• 10: Interpretation of Nominals
• 11: Clauses and Categories
• 12: A Unified Approach to Locality Constraints
• References
• Notes
• Index