Generative Grammar

Geoffrey Horrocks

• Publisher: Routledge
• Year: 2014
• ISBN:
  9781317887775
• Edition: 1
• Topic: Filología
• Subtopic: Lingüística

This book provides a critical review of the development of generative grammar, both transformational and non-transformational, from the early 1960s to the present, and presents contemporary results in the context of an overall evaluation of recent research in the field.Geoffrey Horrocks compares Chomsky's approach to the study of grammar, culminating in Government and Binding theory, with two other theories which are deliberate reactions to this framework: Generalised Phrase Structure Grammar and Lexical-Functional Grammar. Whilst proponents of all three models regard themselves as generative grammarians, and share many of the same objectives, the differences between them nevertheless account for much of the recent debate in this subject. By presenting these different theories in the context of the issues that unite and divide them, the book highlights the problems which arise in any attempt to establish an adequate theory of grammatical representation.

Chapter 1

Aims and assumptions

1.1 Introduction

In the period from the mid-1950s to the mid-1960s Chomsky established a set of objectives for theoretical linguistics which subsequently he himself has consistently sought to attain, and which have continued to direct the course of research for many linguists to the present day. Naturally, during the course of the last thirty years, there have been many modifications, some of them quite radical, to the concrete proposals first put forward in an attempt to achieve these stated goals, and there have also been shifts of focus from one objective to another as work has progressed and previously unanswerable questions have become tractable in the light of new discoveries. But the ultimate goals themselves have changed very little, and it is an essential prerequisite to an understanding of the development of generative grammar to have a clear picture of the aims and philosophical underpinnings of the Chomskyan programme; it is only with these in mind that the full force of the arguments for or against some technical innovation can be appreciated. This is true equally of Chomsky's own revisions and of those advanced by linguists who regard themselves as rivals and critics. It is testimony to Chomsky's prestige in the field, and to the central role that his work has played in its development, that even those generative grammarians who disagree with him tend nevertheless to justify their proposals on the grounds that these achieve the more or less agreed objectives more successfully than his. It is the purpose of this chapter to provide an informal introduction to some of the more important leading ideas of the generative enterprise. Many of the points and issues raised here will be dealt with in more detail and with greater precision in later chapters, where the merits and demerits of various proposals aimed at executing these ideas will be considered.

1.2 Competence and performance

Central to Chomsky's approach to the study of language is a distinction drawn between a native speaker's competence and his performance. The former is defined as the speaker's internalised grammar of his language, his tacit or unconscious knowledge of the system of rules and principles which underlies his capacity to speak and understand the language of his speech community. The latter is the speaker's actual use of language on particular occasions, and includes not only directly observable utterances, spoken and written, but also the speaker's use of language to clarify his thoughts, and other phenomena ‘observable’ only by introspection, such as his ability to pass judgements on the acceptability of utterances in terms of their sound, form and meaning, and his awareness, perhaps partly subconscious, of the existence of various systematic structural and semantic correspondences between certain utterance types, as reflected in his ability to form questions corresponding to statements, passive analogues to active sentences, and so on. The assumption is, then, that the speaker's linguistic capabilities are grounded in, and ultimately explained by, his competence. It is important to note that competence is viewed not as a skill but as a system of knowledge which underlies various skills; it is what the speaker must know in order to be able to perform.

It is evident that performance will only ever be a direct and accurate reflection of competence in ideal circumstances. The utterances produced by speakers tend to have not only properties which are determined by the rules of their internalised grammars, but also properties which derive from factors such as memory limitation, lack of concentration, division of attention, nervousness, inebriation and a whole host of factors which are not, it may be assumed with some plausibility, linguistic in character. It is also often the case that judgements of acceptability are made not instinctively (ie by reference to one's competence alone), but rather self-consciously, say by reference to some rule of good usage taught at school (‘avoid split infinitives’, ‘do not end a sentence with a preposition’, and so on). Equally, an utterance may be rejected not so much on the basis of its linguistic properties as because the hearer cannot readily imagine a context in which the utterance in question could be used appropriately. For example, someone lacking the imagination to construct the sort of context that is typically elaborated in stories for young children might have difficulty in accepting a decontextualised utterance of [1].

[1] The furious carrot slammed the door

The problem, then, is one of determining just which aspects of performance are linguistic in character and so constitute a reflection of the properties of competence. The solution to this problem is not self-evident, and it is obviously possible in principle to draw the line between linguistically relevant and linguistically irrelevant performance phenomena in many different places. As we shall see (3.1, 3.8, 4.1 and 4.9), this question of demarcation is one of the issues which divides different schools of generative grammarians. It should be stressed, however, that this indeterminacy in no way undermines the theoretical value of the distinction itself, to which we now turn.

Despite the indirectness of the relationship between competence and performance, Chomsky believes that the best hope for progress in linguistics is for researchers to concentrate on the elucidation of competence, to concentrate on giving an explicit account of what native speakers know that enables them to perform in the way they do.

A number of important questions arise at this point. The first we have already touched on, namely that of deciding which performance data the theory of competence is to account for. Chomsky's answer to this is unequivocal. The central data which linguistic theory must account for are the introspective intuitions and judgements of native speakers in matters such as grammatical structure, well-formedness, paraphrase relations, ambiguity and so on. But even after we have decided which aspects of performance are to be discounted we must still decide on the basis of what is left just what the unobservable system that underlies this set of capabilities must be like. Can this system be extracted from a sample of the language by mechanical procedure, or must it incorporate principles which are not extractable by any such process? The answer to this question is, not surprisingly, controversial, and Chomsky's view of the matter will be discussed in some detail below (see especially 2.3.11).

The main issue here, however, is the justification of the decision to concentrate on competence when it is clear that there are going to be great difficulties in assessing its character on the basis of the available evidence. Indeed, some linguists, for example those advocating the rival theory of Generalised Phrase Structure Grammar (cf Ch. 3), argue that this available evidence, including not only introspective judgements but also other aspects of performance such as parsing abilities, is not sufficient to justify Chomsky's view that the grammars which linguists construct may plausibly be regarded as models of native speakers' internalised grammars (cf 3.1). For them, linguists' grammars have no real-world interpretation and are to be judged by criteria such as formal precision, generality and elegance of formulation and the extent to which they advance co-operative research ventures with neigh-bouring disciplines. This is important, because it brings to light the fact that what is controversial about Chomsky's position is not that he has decided to restrict and idealise the domain of enquiry but that, having done so, he is still prepared to adopt a realist interpretation of the grammars he constructs. The long and, many would argue, successful Western tradition of grammatical research and theorising gives one grounds for believing that grammar can be studied in isolation in a worthwhile way, and it is, of course, traditional that linguists abstract away from the full range of observable phenomena and concentrate on those aspects which they believe can be satisfactorily handled in a systematic fashion. But it is one thing to do this when no claims are being made about the ‘reality’ of the grammars so constructed (and these artefacts are judged by their utility in some domain such as foreign language teaching) and quite another to do this when claims of psychological reality are involved. Who is to say whether or not in these circumstances the chosen domain of evidence is indeed sufficient to support such a hypothesis? This question is taken up below (1.3).

If for the moment we interpret Chomsky's decision to concentrate on competence simply as a decision to concentrate on grammar, the justification is straightforward. The domain of grammar can be given a reasonably clear preliminary definition that offers the linguist a limited and coherent object of enquiry. That is to say, the factors which influence actual performance in its totality are so varied, and their nature and the manner of their interaction so little understood, that the ordinary creative use of language by native speakers is of necessity going to resist illuminating investigation if it is approached in a monolithic way. It is good methodological sense to adopt a modular approach to the study of language and to deal first with those aspects that seem to lend themselves to systematic treatment. This position only becomes controversial in the context of Chomsky's realism, to which we now turn.

1.3 The interpretation of grammars

Assuming that it is reasonable for any linguist to delimit the domain of fact that his theory seeks to account for, we might nevertheless find fault with Chomsky for asserting that the grammars which his theory sanctions are models of the native speaker's competence. Surely there can never be positive evidence that this or that formulation corresponds to the way in which linguistic knowledge is in fact represented in the native speaker's internalised grammar?

Chomsky's reply to this objection is that it is the condition of research in any subject that is regarded as an empirical discipline that the theories set up to account for the facts that fall within its domain are always in some degree underdetermined by those facts. If this were not so, and if theories could somehow be extracted from samples of data by analysis and generalisation, the theories in question would be nothing more than reduced versions of the facts that the investigator began with and could in no sense be said to explain those facts. In practice, however, most scientific theories are not simply sets of inductive generalisations extracted from a corpus of data; they are rather hypotheses from whose constructs predictions about the behaviour of some set of objects can be deduced and whose reliability can be tested in terms of whether the predicted behaviour matches the observed behaviour. Theories can, of course, be compared in terms of how well the facts follow from the assumptions made. Efficient accounts of the observable phenomena are those which not merely make correct predictions but which do so by employing principles of some generality and predictive power, and furthermore, principles characterised by formal simplicity and compatibility with other theories in related domains. If it is accepted that a science such as physics has been able to progress because scientists have been willing to speculate and advance hypotheses as underlying explanations for observable phenomena, and if one is prepared to accept that such a methodology means that no scientific theory will ever be demonstrably correct, it follows that there can be no concept of ‘physically real’ independent of whatever is normal scientific practice at the time in the field of physics. Obviously it may be the case that the world is constituted in ways which are quite different from those which contemporary physicists propose, but as the principles of modern physics are highly abstract principles of great explanatory power it is reasonable to ask whether this is at all likely.

Relating this to linguistics, it is Chomsky's view that the question of whether his, or anyone else's, theory of competence is the correct one is meaningless. What is psychologically real is whatever the best available theory of competence comprises. Again, it may be the case that the theory in question is fundamentally misconceived, but assuming that linguistics has matured as a discipline to the point there the principles it employs are not merely statements of the superficially obvious but have some degree of generality and predictive power, as Chomsky would argue that it has now begun to do, it is again reasonable to ask whether a theory which provides a penetrating account of the data in its domain could somehow have achieved this degree of success by accident, the system under investigation being ‘in reality’ completely different in its essential properties. Adopting the procedures that are routine in the other sciences, what is needed in linguistics according to Chomsky is a proper set of criteria for evaluating the success of theories of competence.

It is sometimes objected that Chomsky's proposals are essentially formal in character and that the drawing of psychological conclusions is unwarranted. But given that the data of linguistics as conceived by Chomsky (cf 1.4) are psychological in character, it is in his view as ridiculous for a linguist to devise an efficient theory of the linguistic facts but to refuse to regard his theory as having implications for the study of mental faculties as it would be for a physicist, having made careful observations of the behaviour of falling objects, to set up a theory of gravity but to regard this as no more than a convenient fiction that allowed him to talk intelligibly about the behaviour of those objects.

Those who wish to call into question the validity of Chomsky's analogy with the practice of the natural sciences might try to show that the actual practice of linguists revealed a problem of indeterminacy more serious than, and fundamentally different in character from, that faced by, say, physicists. For example, while it is probably true that no one would require independent (ie non-physical) evidence in favour of ascribing physical reality to the formal constructs of theoretical physics, it is not generally regarded as unreasonable to ask for independent evidence (for example from experimental psychology, neurophysiology or whatever) for the psychological reality of the constructs of Chomsky an linguistics. This contrast would perhaps be a telling one if one took the view that Chomsky's characterisation of competence, designed to account for native speakers' introspective judgements over a certain domain, should also be required to account for data in other domains, such as the results of psychological experiments on language processing. If, contrary to Chomsky's view, such an extension of the domain is thought desirable, as indeed it is by proponents of Lexical-Functional Grammar (cf Ch. 4), then the theoretical apparatus proposed would have to be compatible with current theories about the nature of the parsing mechanisms employed by human beings. But it is perfectly legitimate, in the absence of compelling evidence to the effect that the speaker's internalised representation of grammatical knowledge is utilised in a direct way in language processing, simply to deny that observations about language processing have any relevance to the central tasks of theoretical linguistics. The desirability of such a move is discussed in some detail in 4.9. It is, however, worth pointing out here that such a restriction is justified to the extent that it permits the execution of a fruitful research strategy. One is not, after all, very surprised to discover that the physicists' laws of motion do not actually predict with any accuracy the movements of one's pet dog at any given moment; indeed, if physicists had set themselves the task of accounting for the movements of all animate beings in the universe over and above the tasks normally regarded as falling within the domain of their science, there is good reason to think that physics would be a much less successful science today than in fact it is. The point being made is simply that the inclusion of ‘independent’ evidence within the domain of theoretical linguistics as conceived by Chomsky is not necessarily going to be very helpful. The only way to judge the value of such a move ultimately is to construct theories that seek to account for this broader range of data and which presuppose direct use of the grammatical knowledge-store in language processing, and to see if this proves to be a fruitful line of enquiry. To the extent that lexical-functional grammars have succeeded in unifying research in linguistics and experimental psychology they constitute a serious challenge to Chomsky's stance. They do not, however, show that Chomsky's position is ‘wrong’ in any absolute sense, since it is undeniable that many very valuable insights have come out of research conducted within the more restricted domain that he advocates (see in particular 2.3).

All this said, it is still true that in linguistics there is considerably more room for disagreement about what constitutes a reasonable analysis or interpretation of a given range of data than would ordinarily be the cas...

Table of contents

• Cover
• Half Title
• Longman Linguistics Library
• Title Page
• Copyright Page
• Table of Contents
• Preface
• 1 Aims and assumptions
• 2 Chomsky's theory of grammar
• 3 Generalised Phrase Structure Grammar
• 4 Lexical-Functional Grammar
• 5 The theories compared
• Bibliography
• Index