As the name indicates, universal grammar was concerned with general features of language structure rather than with particular idiosyncrasies. Particularly in France, universal grammar developed in part in reaction to an earlier descriptivist tradition which held that the only proper task for the grammarian was to present data, to give a kind of ‘natural history’ of language (specifically, of the ‘cultivated usage’ of the court and the best writers). In contrast, universal grammarians urged that the study of language should be elevated from the level of ‘natural history’ to that of ‘natural philosophy’; hence the term ‘philosophical grammar’, ‘philosophical’ being used, of course, in essentially the sense of our term ‘scientific’. Grammar should not be merely a record of the data of usage, but, rather, should offer an explanation for such data. It should establish general principles, applicable to all languages and based ultimately on intrinsic properties of the mind, which would explain how language is used and why it has the particular properties to which the descriptive grammarian chooses, irrationally, to restrict his attention.

Universal grammarians did not content themselves with merely stating this goal. In fact, many generations of scholars proceeded to develop a rich and far-reaching account of the general principles of language structure, supported by whatever detailed evidence they could find from the linguistic materials available to them. On the basis of these principles, they attempted to explain many particular facts, and to develop a psychological theory dealing with certain aspects of language use, with the production and comprehension of sentences.

The tradition of universal grammar came to an abrupt end in the nineteenth century, for reasons that I will discuss directly. Furthermore, its achievements were very rapidly forgotten, and an interesting mythology developed concerning its limitations and excesses. It has now become something of a cliché among linguists that universal grammar suffered from the following defects: (a) it was not concerned with the sounds of speech, but only with writing; (b) it was based primarily on a Latin model, and was, in some sense ‘prescriptive’; (c) its assumptions about language structure have been refuted by modern ‘anthropological linguistics’. In addition, many linguists, though not all, would hold that universal grammar was misguided in principle in its attempt to provide explanations rather than mere description of usage, the latter being all that can be contemplated by the ‘sober scientist’.

The first two criticisms are quite easy to refute; the third and fourth are more interesting. Even a cursory glance at the texts will show that phonetics was a major concern of universal grammarians, and that their phonetic theories were not very different from our own. Nor have I been able to discover any confusion of speech and writing. The belief that universal grammar was based on a Latin model is rather curious. In fact, the earliest studies of universal grammar, in France, were a part of the movement to raise the status of the vernacular, and are concerned with details of French that often do not even have any Latin analogue.

As to the belief that modern ‘anthropological linguistics’ has refuted the assumptions of universal grammar, this is not only untrue, but, for a rather important reason, could not be true. The reason is that universal grammar made a sharp distinction between what we may call ‘deep structure’ and ‘surface structure’. The deep structure of a sentence is the abstract underlying form which determines the meaning of the sentence; it is present in the mind but not necessarily represented directly in the physical signal. The surface structure of a sentence is the actual organization of the physical signal into phrases of varying size, into words of various categories, with certain particles, inflections, arrangement, and so on. The fundamental assumption of the universal grammarians was that languages scarcely differ at the level of deep structure—which reflects the basic properties of thought and conception—but that they may vary widely at the much less interesting level of surface structure. But modern anthropological linguistics does not attempt to deal with deep structure and its relations to surface structure. Rather, its attention is limited to surface structure—to the phonetic form of an utterance and its organization into units of varying size. Consequently, the information that it provides has no direct bearing on the hypotheses concerning deep structure postulated by the universal grammarians. And, in fact, it seems to me that what information is now available to us suggests not that they went too far in assuming universality of underlying structure, but that they may have been much too cautious and restrained in what they proposed.

The fourth criticism of universal grammar—namely, that it was misguided in seeking explanations in the first place—I will not discuss. It seems to me that this criticism is based on a misunderstanding of the nature of all rational inquiry. There is particular irony in the fact that this criticism should be advanced with the avowed intention of making linguistics ‘scientific’. It is hardly open to question that the natural sciences are concerned precisely with the problem of explaining phenomena, and have little use for accurate description that is unrelated to problems of explanation.

I think that we have much to learn from a careful study of what was achieved by the universal grammarians of the seventeenth and eighteenth centuries. It seems to me, in fact, that contemporary linguistics would do well to take their concept of language as a point of departure for current work. Not only do they make a fairly clear and well-founded distinction between deep and surface structure, but they also go on to study the nature of deep structure and to provide valuable hints and insights concerning the rules that relate the abstract underlying mental structures to surface form, the rules that we would now call ‘grammatical transformations’. What is more, universal grammar developed as part of a general philosophical tradition that provided deep and important insights, also largely forgotten, into the use and acquisition of language, and, furthermore, into problems of perception and acquisition of knowledge in general. These insights can be exploited and developed. The idea that the study of language should proceed within the framework of what we might nowadays call ‘cognitive psychology’ is sound. There is much truth in the traditional view that language provides the most effective means for studying the nature and mechanisms of the human mind, and that only within this context can we perceive the larger issues that determine the directions in which the study of language should develop.

The tradition of universal grammar came to an end more than a century ago. Several factors combined to lead to its decline. For one thing, the problems posed were beyond the scope of the technique and understanding then available. The problem of formulating the rules that determine deep structures and relate them to surface structures, and the deeper problem of determining the general abstract characteristics of these rules, could not be studied with any precision, and discussion therefore remained at the level of hints, examples, and vaguely formulated intentions. In particular, the problem of rule-governed creativity in language simply could not be formulated with sufficient precision to permit research to proceed very far. A second reason for the decline of traditional linguistic theory lies in the remarkable successes of Indo-European comparative linguistics in the nineteenth century. These achievements appeared to dwarf the accomplishments of universal grammar, and led many linguists to scoff at the ‘metaphysical’ and ‘airy pronouncements’ of those who were attempting to deal with a much wider range of problems—and at that particular stage of the development of linguistic theory, were discussing these topics in a highly inconclusive fashion. Looking back now, we can see quite clearly that the concept of language employed by the Indo-European comparativists was an extremely primitive one. It was, however, well-suited to the tasks at hand. It is, therefore, not too surprising that this concept of language, which was then extended and developed by the structural and descriptive linguists of the twentienth century, became almost completely dominant, and that the older tradition of linguistic theory was largely swept aside and forgotten. This is hardly a unique instance in intellectual history.

Structural linguistics is a direct outgrowth of the concepts that emerged in Indo-European comparative study, which was primarily concerned with language as a system of phonological units that undergo systematic modification in phonetically determined contexts. Structural linguistics reinterpreted this concept for a fixed state of a language, investigated the relations among such units and the patterns they form, and attempted with varying success, to extend the same kind of analysis to ‘higher levels’ of linguistic structure. Its fundamental assumption is that procedures of segmentation and classification, applied to data in a systematic way, can isolate and identify all types of elements that function in a particular language along with the constraints that they obey. A catalogue of these elements, their relations, and their restrictions of ‘distribution’, would, in most structuralist views, constitute a full grammar of the language.

Structural linguistics has very real accomplishments to its credit. To me, it seems that its major achievement is to have provided a factual and a methodological basis that makes it possible to return to the problems that occupied the traditional universal grammarians with some hope of extending and deepening their theory of language structure and language use. Modern descriptive linguistics has enormously enriched the range of factual material available, and has provided entirely new standards of clarity and objectivity. Given this advance in precision and objectivity, it becomes possible to return, with new hope for success, to the problem of constructing the theory of a particular language—its grammar—and to the still more ambitious study of the general theory of language. On the other hand, it seems to me that the substantive contributions to the theory of language structure are few, and that, to a large extent, the concepts of modern linguistics constitute a retrogression as compared with universal grammar. One real advance has been in universal phonetics—I refer here particularly to the work of Jakobson. Other new and important insights might also be cited. But in general, the major contributions of structural linguistics seem to me to be methodological rather than substantive. These methodological contributions are not limited to a raising of the standards of precision. In a more subtle way, the idea that language can be studied as a formal system, a notion which is developed with force and effectiveness in the work of Harris and Hockett, is of particular significance. It is, in fact, this general insight and the techniques that emerged as it developed that have made it possible, in the last few years, to approach the traditional problems once again. Specifically, it is now possible to study the problem of rule-governed creativity in natural language, the problem of constructing grammars that explicitly generate deep and surface structures and express the relations between them, and the deeper problem of determining the universal conditions that limit the form and organization of rules in the grammar of a human language. When these problems are clearly formulated and studied, we are led to a conception of language not unlike that suggested in universal grammar. Furthermore, I think that we are led to conclusions regarding mental processes of very much the sort that were developed, with care and insight, in the rationalist philosophy of mind that provided the intellectual background for universal grammar. It is in this sense that I think we can look forward to a productive synthesis of the two major traditions of linguistic research.

If this point of view is correct in essentials, we can proceed to outline the problems facing the linguist in the following way. He is, first of all, concerned to report data accurately. What is less obvious, but nonetheless correct, is that the data will not be of particular interest to him in itself, but rather only insofar as it sheds light on the grammar of the language from which it is drawn, where by the ‘grammar of a language’ I mean the theory that deals with the mechanisms of sentence construction, which establish a sound-meaning relation in this language. At the next level of study, the linguist is concerned to give a factually accurate formulation of this grammar, that is, a correct formulation of the rules that generate deep and surface structures and interrelate them, and the rules that give a phonetic interpretation of surface structures and a semantic interpretation of deep structures. But, once again, this correct statement of the grammatical principles of a language is not primarily of interest in itself, but only insofar as it sheds light on the more general question of the nature of language; that is, the nature of universal grammar. The primary interest of a correct grammar is that it provides the basis for substantiating or refuting a general theory of linguistic structure which establishes general principles concerning the form of grammar.

Continuing one step higher in level of abstraction, a universal grammar—a general theory of linguistic structure that determines the form of grammar—is primarily of interest for the information it provides concerning innate intellectual structure. Specifically, a general theory of this sort itself must provide a hypothesis concerning innate intellectual structure of sufficient richness to account for the fact that the child acquires a given grammar on the basis of the data available to him. More generally, both a grammar of a particular language and a general theory of language are of interest primarily because of the insight they provide concerning the nature of mental processes, the mechanisms of perception and production, and the mechanisms by which knowledge is acquired. There can be little doubt that both specific theories of particular languages and the general theory of linguistic structure provide very relevant evidence for anyone concerned with these matters; to me it seems quite obvious that it is within this general framework that linguistic research finds its intellectual justification.

Performance provides evidence for the investigation of competence. At the same time, a primary interest in competence entails no disregard for the facts of performance and the problem of explaining these facts. On the contrary, it is difficult to see how performance can be seriously studied except on the basis of an explicit theory of the competence that underlies it, and, in fact, contributions to the understanding of performance have largely been by-products of the study of grammars that represent competence.

Notice, incidentally, that a person is not generally aware of the rules that govern sentence-interpretation in the language that he knows; nor, in fact, is there any reason to suppose that the rules can be brought to consciousness. Furthermore, there is no reason to expect him to be fully aware even of the empirical consequences of these internalized rules—that is, of the way in which signals are assigned semantic interpretations by the rules of the language that he knows (and, by definition, knows perfectly). It is important to realize that there is no paradox in this; in fact, it is precisely what should be expected.

Current work in generative grammar has adopted this traditional framework of interests and concerns. It attempts to go beyond traditional grammar in a fundamental way, however. As has repeatedly been emphasized, traditional grammars make an essential appeal to the intelligence of the reader. They do not actually formulate the rules of the grammar, but rather give examples and hints that enable the intelligent reader to determine the grammar, in some way that is not at all understood. They do not provide an analysis of the ‘faculté de langage’ that makes this achievement possible. To carry the study of language beyond its traditional bounds, it is necessary to recognize this limitation and to develop means to transcend it. This is the fundamental problem to which all work in generative grammar has been addressed.

The most striking aspect of linguistic competence is what we may call the ‘creativity of language’, that is, the speaker’s ability to produce new sentences, sentences that are immediately understood by other speakers although they bear no physical resemblance to sentences which are ‘familiar’. The fundamental importance of this creative aspect of normal language use has been recognized since the seventeenth century at least, and it was at the core of Humboldtian general linguistics. Modern linguistics, however, is seriously at fault in its failure to come to grips with this central problem. In fact, even to speak of the hearer’s ‘familiarity with sentences’ is an absurdity. Normal use of language involves the production and interpretation of sentences that are similar to sentences that have been heard before only in that they...