GT-1

The extraordinary and traumatic impact of the publication of Syntactic Structures by Noam Chomsky in 1957 can hardly be appreciated by one who did not live through this upheaval.¹ Chomsky denies the fundamental assumption of structuralism by arguing that an adequate linguistic description of grammar cannot be derived by applying sets of operations to primary data but rather must be viewed as a formal deductive theory whose object is to separate the grammatical sentences of a language from the ungrammatical ones and to provide a systematic account of the structure of grammatical sentences. This position involves a complete reversal of the relations among the parts of a linguistic description (as symbolized by the direction of arrows in Figure 2 as opposed to Figure 1) as well as defining the primary object of linguistic theory to be the principles underlying the construction of sentences rather than the identification of minimal signaling units such as phonemes.

Chomsky’s work has led to a genuine scientific revolution in that his approach has redefined the goals and methods of linguistics and thereby delineated a set of relevant problems with which linguists may be properly concerned. Sklar’s (1968) account of Chomsky’s efforts to develop a rigorous analysis of syntax based on structuralist principles and his conclusion that a multitude of stubborn linguistic facts simply could not be handled by operational procedures is an illustration of Kuhn’s (1962) description of the way in which a paradigm, during a period of normal problem solving science, accumulates implicit counterexamples which are, at some point, made explicit leading to the development of a new paradigm.

Once one becomes seriously interested in syntax (a natural consequence of being interested in language), procedures which seem to work effectively on phonological and morphological problems are no longer satisfactory. Since the lexicon of a language must be finite, it may seem at least possible in principle to enumerate all of the environments relevant to the identification of phonemes. One’s confidence is likely to ebb rapidly when considering the possibilities of a similar analysis for the infinite array of sentences apart from the many problems of correctly stating the internal structure of sentences and the relationships among them.

The next step is to conclude that the procedures didn’t really work very well in the simpler cases either as the problems on the other levels are seen in a new light.

Chomsky’s success may be associated with some particular characteristics of linguistics as a discipline as well as with the merits of his position. The fact that human languages had been an object of serious study by many scholars for centuries had created a situation where there was essential agreement on a large body of empirical facts about language and many methods of analysis had been investigated. It was thus possible to measure a new proposal against a rather well developed empirical domain and Chomsky’s impact is due in no small part to his ability to offer solutions to a wide range of problems that had been either ignored or handled clumsily by structural methods. One cannot imagine any theorist having had such a rapid and dramatic effect in anthropology or sociology.

In addition to the relatively advanced state of linguistic research, the social organization of academic linguistics, especially in the United States, provided (and still provides) an environment where a new idea could enjoy rapid dissemination. There were only a few universities which offered advanced degrees in linguistics and perhaps no more than three major journals (Language, Word, and The International Journal of American Linguistics) in which purely linguistic research regularly appeared. The Linguistic Society of America was a comparatively small academic group whose members came into frequent personal contact both through winter and summer meetings and the regular summer institutes sponsored by the society at various universities. The traditional policy of presenting only one sequence of papers to the whole society at meetings along with the convention that all papers are open to pointed comment from the floor guaranteed that a new viewpoint would be both presented and publicly debated.² Although Chomsky himself was not often an active participant in these affairs, his views were vigorously presented during this early period by his associates, especially Robert B. Lees. One interesting correlative of the rise of transformational grammar to a dominant position has been the striking quantitative expansion of linguistics in American universities. Although a number of general social factors are surely involved in this development, one causal element may well be the increased power and relevance of linguistic theory.

The sure sign of impending revolutionary success is the conversion of the young, and this occurred rapidly as graduate students became transformationalists while their own professors continued to profess structuralism. In addition, many superior students and younger scholars entered linguistics from such fields as philosophy and mathematics. That the originators of paradigms are not immune to such developments will be seen in the discussion of GT-3 where a similar process of revolt seems to be occurring. Just as Chomsky came to disagree with many of the teachings of Zelig Harris, so many of his own students now resist his views. It is, of course, necessary in both cases to emphasize the evolutionary nature of change as well as the more striking revolutionary aspects. At least in linguistics there is a marked continuity of development across the positions described here. I will attempt to show that while Chomsky’s innovations in syntax were truly radical, he did retain the central structuralist assumption as to the independence of form and meaning. Further, although his view of meaning differed in a number of important respects from that of Bloomfield and Harris, the full implications of this change were not exploited during this period and, in effect, the status of meaning in a linguistic description remained much the same.

The machinery of a linguistic analysis is now seen as consisting of a set of rules whose goal is to generate, automatically, all and only the grammatical sentences of a language along with a structural description for each which shows how its parts are combined to form the full sentence. Two types of rules define the two major components of a grammar.³

(1) Phrase structure (or constituent structure) rules are of the form XAY-> XBY where A is a single symbol, B is a string of symbols and the environment within which the rule applies consists of X_Y which may be omitted if the rule is not subject to environmental constraints (i.e. is ‘context free’). Consider the fragment of English defined by the following set of phrase structure rules:

These rules will permit the derivation of such strings as The + boys + # like + girls, Girls + # + like + rabbits, etc. and automically assign a structure to each:

(2) Transformational rules are of the form P ==> Q where P and Q are strings of symbols and where the rewriting rule may rearrange the symbols of P, delete some of them, or add new symbols, to form Q (simple or singulary transformations). In addition, these rules may combine two strings to form a third: P₁/P₂ ==> Q (generalized transformations).

Transformations apply to labeled terminal strings of the phrase structure component in Syntactic Structures. In subsequent work (e.g. Chomsky and Miller, 1963) transformations are seen as converting P-markers (P) into a new P-Marker (Q). A P-marker may be regarded as equivalent to a labeled tree as above. For convenience, we will adopt the representation of transformations given in Syntactic Structures. Applied to the output of the phrase structure component these rules will produce more complex sentences (or, more properly, the strings underlying them) and state formally the relationship between transformationally derived sentences and the more basic strings from which they are derived.

The output of the phrase structure component will not consist of perfectly well formed strings. In the example above, # is not yet interpreted. Given the constraint that phrase structure rules cannot be of the form A -----> ɸ and that they must depend on immediate environments, we require the following two obligatory transformations in order to achieve number agreement

These will produce kernel strings such as:

The essential substantive division here is between the kernel of a language consisting of simple active declarative sentences produced by the phrase structure rules and the obligatory transformations and the other more complex sentences formed by applying optional transformations to the set of kernel sentences. Simplified versions of the Relative and Question transformations will illustrate the way in which complex sentences are formed from simple ones:

The morphophonemic component of Chomsky (1957) and Lees (i960) contains rules of the phrase structure type which convert the well formed strings of morphemes produced by the transformational component into a phonemic representation. Thus, the S in the rules above is converted into the proper phonemic form (/-s/, /-0z/, or /-z/) by a straightforward statement of the phonemic environments in which these forms occur. This formulation is of interest here because of its similarities in terminology and content to the phonological component of a structural grammar. It was soon replaced by a full-fledged generative phonological component which applied transformational types of rules to the output of the syntactic component in order to produce phonetic representations.

This approach to language both restricts the goals of linguistic analysis and shifts the domain of linguistic theory. The structuralist goal of developing an automatic discovery procedure is rejected in favor of the more modest aim of providing an apparatus for the evaluation of competing grammars of a given language. The domain of linguistic theory becomes the knowledge that native speakers have about the formal properties of their language rather than the set of observable physical utterances which they produce. This has the effect of directing the attention of linguists toward the inner mental equipment of speakers at the expense of a direct concern with their overt behavior. Not very much is made of the distinction between competence and performance in Syntactic Structures although references to it are frequent in subsequent work (e.g. Chomsky and Miller, 1963). One of the most striking substantive contrasts with structuralism is the necessity of viewing human language as a phenomenon of immense complexity. This had a reverberating effect in the related areas of language acquisition and use where simplistic models based on behavior theory came under severe attack.⁴ It is correct to say that the complexity of language, as viewed by linguists, has been steadily increasing ever since.

Bloomfield’s justification for excluding meaning from linguistics rested on his definition of the term which made an account of meaning equivalent to an account of the total social, cultural and individual context of speech. In order for linguistics to have any hope of handling semantics this global definition must be partitioned into a part which is narrowly ‘ linguistic ’ as supposed to a part which includes the remainder of human knowledge. Chomsky’s rejection of the structuralist definition provides the basis for such a distinction:

It is strange that those who have objected to basing linguistic theory on such formulations as (117 i) * should have been accused of disregard for meaning. It appears to be the case, on the contrary, that those who propose some variant of (117 i) must be interpreting ‘meaning’ so broadly that any response to language is called ‘meaning’. But to accept this view is to denude the term ‘ meaning ’ of any interest or significance. I think that anyone who wishes to save the phrase ‘ study of meaning ’ as descriptive of an important aspect of linguistic research must reject this identification of ‘ meaning ’ with ‘response to language’, and along with it, such formulations as (117!). (Syntactic Structures, 99-100)

* 117 i is the assertion that two utterances are phonemically distinct if and only if they differ in meaning.

At the same time, Harris’ rejection of meaning as a basis for formal description is retained by Chomsky:

It is, of course, impossible to prove that semantic notions are of no use in grammar, just as it is impossible to prove the irrelevance of any other given set of notions. Investigation of such proposals, however, invariably seems to lead to the conclusion that only a purely formal basis can provide a firm and productive foundation for the construction of grammatical theory. (Syntactic Structures, 100)

Like Bloomfield and Harris, Chomsky notes the many correlations between form and meaning but, unlike the structuralists, proposes that semantic factors may function as important criteria for the evaluation of grammars:

The fact that correspondences between formal and semantic features exist, however, cannot be ignored. These correspondences should be studied in some more general theory of language that will include a theory of linguistic form and a theory of the use of language as subparts. In §8 we found that there are, apparently, fairly general types of relations between these two domains that deserve more intensive study. Having determined the syntactic structure of the language, we can study the way in which the syntactic structure is put to use in the actual functioning of language. An investigation of the semantic function of level structure, as suggested briefly in §8, might be a reasonable step towards a theory of the interconnections between syntax and semantics. In fact, we pointed out in § 8 that the correlations between the form and use of language can even provide certain rough criteria of adequacy for a linguistic theory and the grammars to which it leads. We can judge formal theories in terms of their ability to explain and clarify a variety of facts about the way in which sentences are used and understood. In other words, we should like the syntactic framework of the language that is isolated and exhibited by the grammar to be able to support semantic description, and we shall naturally rate more highly a theory of formal structure that leads to grammars that meet this requirement more fully. (Syntactic Structures, 102)

It will be noted that, anticipating the competence-performance distinction, meaning is classified as performance (‘use’). The subsequent association of semantic interpretation with ‘ deep structure ’ is also foreshadowed:

To understand a sentence we must know much more than the analysis of this sentence on each linguistic level. We must also know the reference and meaning of the morphemes or words of which it is composed. These notions form the subject matter for semantics. In describing the meaning of a word it is often expedient, or necessary, to refer to the syntactic framework in which this word is usually embedded; e.g., in describing the meaning of ‘ hit ’ we would no doubt describe the agent and object of the action terms of the notions ‘ subject ’ and ‘ object ’, which are apparently best analyzed as purely formal notions belonging to the theory of grammar.* (Syntactic Structures, 104)

* Such a description of the meaning of ‘ hit’ would then account automatically for the use of ‘ hit ’ in such transforms as ‘ Bill was hit by John,’ ‘ hitting Bill was wrong ’, etc., if we can show in sufficient detail and generality that transforms are ‘ understood ’ in terms of the underlying kernel sentences.

In effect, the status of meaning with regard to formal description is much like that found in structuralist accounts. It is outside of linguistics proper and clearly secondary to the description of syntax. However, the attempt to distinguish ‘ linguistic meaning ’ and the proposal that grammars are to be more highly evaluated to the extent that they support a valid account of the way in which sentences are ‘ understood ’ provides a rationale for the systematic incorporation of meaning into linguistic description. Indeed it requires such a development. If grammars must only distinguish well-formed from ill-formed sentences and provide structural descriptions for the well-formed sentences as stated in Chomsky’s discussion of the goals of linguistics, then questions of meaning, however it is defined, may be relegated to some non-linguistic area such as ‘ use ’ or ‘ performance ’. But once meaning is accepted as a criterion for the evaluation of grammars, the necessity of systematically describing it becomes clear. Just as an interest in language leads naturally to a concern with syntax, so a commitment to syntax requires that some attention be given to meaning with consequences that are often not obvious at the beginning of such a study.

¹ An especially good account of Chomsky’s views is given in Lees (1957) while Lees (i960) represents the first major application of the theory of Syntactic Structures to a natural language.

² As a result of a greatly expanded membership the Linguistic Society has now begun to present papers in parallel sessions at its winter and summer meetings although specialization has not yet gotten to the point of organizing sessions around some relatively narrow theme.

³ See Bach (1964), chapters 3 and 4, for a good discussion of these rule types.

⁴ See Chomsky (1959) and Miller, Galanter and Pribram (i960). Maclay (1964) describes the general effect of Chomsky’s views on psycholinguistic research.