Pragmatics and Semantics

An Empiricist Theory

Carol A. Kates

• Publisher: Cornell University Press
• Year: 2020
• ISBN:
  9781501752186
• Topic: Langues et linguistique
• Subtopic: Sémantique linguistique

What is the nature of communicative competence? Carol Kates addresses this crucial linguistic question, examining and finally rejecting the rationalistic theory proposed by Noam Chomsky and elaborated by Jerrold J. Katz, among others. She sets forth three reasons why the rationalistic model shoudl be rejected: (1) it has not been supported by empirical tests; (2) it cannot accommodate the pragmatic relation between speaker and sign; and (3) the theory of universal grammar carries with it unacceptable metaphysical implications unless it is interpreted in light of empiricism. Kates proposes an empiricist model in place of the rationalistic theory—a model that, in her view, is more consistent with recent findings in linguistics and psycholinguistics.

In attempting to clarify the nature of utterance meaning, Kates develops theoretical perspectives on phenomenological empiricism and produces an account of reference and intentionality directly relevant to empiricaly based theories of speaking and understanding.

Among the major topics addressed in the book are transformational-generative and universal grammer, cognitive theories of language acquisition, pragmatic structure, predication and topic-comment structure, and empiricism and the philosophical problem of universals.

An innovative and probing work, Pragmatics and Semantics will be welcomed by philosophers, linguists, and psycholinguists.

PART I

PRAGMATICS AND SEMANTICS: THE NATURE OF COMMUNICATIVE COMPETENCE

CHAPTER 1

Transformational-Generative Grammar

The study of language—its structure, acquisition, and influence on perception and cognition—is, and for several years has been, a vital area of interdisciplinary research. The so-called Chomskyan revolution in linguistics created a somewhat virulent but exciting controversy between transformationalists and distributionalists, which focused attention on language as a system only partially expressed through an actual speech corpus. Speech as an observable behavior needs to be explained by a theory of linguistic competence, describing what a speaker knows about the language.

Distributional grammars focused on regularities in speech which seemed to be productive within a given community of speakers, and thus were structures of the language at a given stage of historical development. A grammar was empirically justified if it presented a somewhat idealized representation of structures abstracted from a corpus of utterances acceptable to the speakers of the community. The distributional grammars written by such pioneers as Leonard Bloomfield (1933) described phonological and grammatical (syntactic and morphological) structures, avoiding or at least postponing the investigation of semantic and pragmatic structure in language. Bloomfield was a behaviorist who did not wish to posit any unobservable mental entities that might endanger the objectivity of structural analysis. The alternative to the sort of mentalistic semantic theory he rejected was, he thought, a behaviorist account; he defined the meaning of a linguistic form as “the situation in which the speaker utters it and the response which it calls forth in the hearer” (p. 139). Unfortunately, such an account of meaning effectively ruled out the possibility of an adequate, scientific description of semantic structure. In Bloomfield’s words: “The situations which prompt people to utter speech, include every object and happening in their universe. In order to give a scientifically accurate definition of meaning for every form of language, we should have to have scientifically accurate knowledge of everything in the speakers’ world. The actual extent of human knowledge is very small, compared to this” (p. 139). Bloomfield concluded that semantics was the “weak point” in language study, and would remain so until general human knowledge made a considerable advance.

Despite the absence of a strictly verifiable semantic theory, linguists were constrained to use semantic clues in reconstructing the grammatical system of a language. Bloomfield acknowledged that linguists had to use such makeshift devices as ostensive definition, circumlocution, and translation into a second language to elicit informants’ judgments about the name of some referent or the equivalence of some expressions, or to discover what semantic function was served by a particular grammatical construction. Bloomfield’s behavioristic model of semantics has been restated more recently by Charles F. Hockett (1958), who defined linguistic meanings as associative ties between morphemes and morpheme-combinations and types of things and situations in the world.¹ In Hockett’s view, it is not the business of linguistics to investigate or describe these associations. However, since they will be “more or less the same for all speakers of a language” (p. 139), the linguist can make a fairly accurate use of semantic criteria to discover the distributional structure of a language.

Distributional grammars might be said to provide a model of competence in the sense that they describe the most productive patterns in the speech of a given community of individuals. Presumably, regularities in speech reflect the grammatical as well as semantic structures that have been internalized by speakers as the result of exposure to a particular language environment. Of course, speakers do not have a conscious knowledge of the distributional structure of their language. Nonetheless, they have acquired, possibly through some complex form of associative learning, a stock of linguistic forms which may be used habitually, in certain types of situations, to communicate certain intentions. On this account, it is the task of the psychologist to provide the learning theory and that of the linguist to describe the speech patterns that reflect the most widely developed linguistic habits.

Chomsky’s (1959) review of Skinner’s Verbal Behavior was the beginning of an argument against distributional grammar and the behaviorist psychology on which it seemed to rest.² Chomsky argued that linguistics must provide a theory of competence that accounts for the creativity of speech. It is an essential fact about language that speakers are able to produce and understand novel but appropriate and acceptable utterances, and it is precisely this fact, he claimed, that behaviorism cannot explain. If competence were simply a matter of associative learning or habit formation, then speakers could never produce or understand a previously unheard sentence. Chomsky echoes Descartes in his argument that the obvious independence of most utterances from any definable set of stimuli and the unpredictability and creativity of speech are clear evidence of the autonomy of the human mind. Thus Chomsky proposes a “mentalistic” and rationalistic theory of transformational-generative (T-G) grammar which he claims will provide an adequate model of competence.

I believe that to evaluate Chomsky’s criticism of empiricism one must distinguish between an empiricist learning theory in a general sense, and, for example, behaviorism as a special type of learning theory. An empiricist might agree that there may never be an adequate behaviorist model of language acquisition (for example, a Skinnerian model characterizing language learning as a type of operant conditioning, defining the reinforcers for speech, and discovering stimulus conditions that control “verbal behavior”). Certainly there are at this time no convincing behaviorist theories of language acquisition. Further, an adequate model of semantic structures may in fact involve mentalistic cognitive structures. Behaviorism, however, is not the only possible type of empiricist approach to language. An empiricist model would require only that there be a fixed connection between linguistic forms and patterns and experiential schemas such that the latter provide an adequate foundation for the creativity exhibited by speakers of a language. The question of how experience shapes linguistic competence can be resolved only through the empirical study of language acquisition and, more generally, of the speech process.

Bloomfield and Hockett, among others, have attributed linguistic creativity to a process of analogical construction (Bloomfield, 1933, p. 275; Hockett, 1958, pp. 356–57, 425–26). For example, if one understands the terms and the construction involved in the sentence “Mary is riding a horse,” and if one knows the meaning of ‘tame’ and ‘tiger’, one can easily produce the (previously unheard) sentence, “Mary is riding a tame tiger.” Similarly, if one grasps the meaning of the plural suffix ‘-s’, one can easily fill in the missing term in the following analogy: ‘chair’ : ‘chairs’ :: ‘table’ : ‘_____’. The difficult point is to describe the basic paradigms or productive patterns of the language and to explain how they are acquired.

Chomsky has rejected this sort of empiricist account of creativity, because, in his words, the concepts of habit or analogy are used “in a completely metaphorical way, with no clear sense and with no relation to the technical usage of linguistic theory” (1966b, p. 12).³ In contrast to the “vague” notion of analogy, then, Chomsky wishes to introduce the “precise,” explicit, and technical notion of an abstract grammatical rule. Speakers who know the rules of grammar are competent to produce and understand any possible sentence of a given language.

According to Chomsky, the rules of grammar are of two types: “phrase structure” rules and “transformational” rules. The former are “rewrite” rules⁴ that assign the syntactic “deep structure” appropriate to some sentence. For example, a simplified phrase structure grammar might include the following sequence of rules: S (sentence) → NP (noun phrase) + VP (verb phrase); NP → Det. (determiner) + N (noun, singular or plural); VP → Verb + NP, and so on, ending with “terminal strings” of words and morphemes of the appropriate syntactic category. The strings, or sequences of symbols, which repeated application of these rules produces are called phrase markers, or hierarchically structured, labeled bracketings conventionally represented by tree diagrams. These deep structure phrase markers provide syntactic information essential to the semantic interpretation of sentences.⁵

The deep structures that are the “output” of the phrase structure component of the grammar may, in turn, undergo one or more transformations before emerging as “surface structures”: the strings of words and morphemes which constitute the set of acceptable sentences of a language. Transformational rules describe a set of operations, such as deletion, embedding, permutation, and so on, that “transform” the abstract deep structures generated⁶ by phrase structure rules. For example, “Mary hit the ball” and “The ball was hit by Mary” have the same deep structure, but the passive form has undergone a transformation.⁷

Chomsky characterizes a grammar, in the broadest sense, as having three components: a syntactic component, including phrase structure and transformational rules; a phonological component, which provides a phonological interpretation of morphemes (converting surface structures into a phonetic representation); and a semantic component, which provides a semantic interpretation on the basis of (deep) syntactic information. Because of Chomsky’s original claim that the semantic component of grammar serves an interpretive function in relation to the deep structures provided by the syntactic component, Chomsky’s version of T-G theory has been dubbed “interpretive semantic” grammar, in contrast to the “generative semantic” position I shall describe in Chapter 2.

Chomsky has proposed several versions of T-G grammar and has recently changed his mind about one difficult aspect of the relation between syntax and semantics. In the model he presented in Aspects of the Theory of Syntax (1965), Chomsky argued that semantic interpretation of sentences operates on the syntactic deep structures generated by the phrase structure or “base” component of the grammar, and that semantic interpretation is not affected by transformations. His proposal was that semantic interpretation is not based on surface structure and that an adequate deep structure analysis of a sentence should provide sufficient information for interpreting the sentence, for specifying multiple interpretations when it is ambiguous, and for showing that it is equivalent to some other sentence whenever they have the same deep structure. As a case in point, consider Chomsky’s most famous example: “Flying planes can be dangerous.” If one simply assigns phrase markers to (or performs an immediate constituent analysis of) this surface structure, one gets the following hierarchical structure:

( ( (flying) (planes) ) ( ( (can) (be) ) (dangerous) ) )

But this sort of analysis does not differentiate the two possible meanings of the sentence and thus does not provide an adequate basis for semantic interpretation. What does show the difference, according to Chomsky, is a transformational analysis that derives the same surface structure from two different deep structures. In one case, phrase structure rules are applied to the kernel sentence: “Planes which fly [or, “are flying”] can be dangerous,” and in the other case to the kernel sentence: “To fly planes can be dangerous.” One interprets the ambiguous surface structure by analyzing its possible deep structures.

Chomsky (1975) has modified his theory of semantic interpretation in response to a number of counterexamples that seem to show that such interpretations are affected by transformations.⁸ The more recent “trace” theory posits an “enriched” surface structure, which retains “traces” of what were formerly called deep structure interpretations, as the object of semantic interpretation (pp. 81–82, 116–17). Chomsky has also decided to drop the term “deep structure” in favor of “initial [i.e., first stage] phrase marker,” first because the former expression (falsely) implied to many that the other aspects of grammar are somehow “superficial, unimportant, and variable across languages” and, second, because, in the standard theory, deep structures initiated transformational derivations and determined semantic interpretations. At this point, Chomsky no longer believes that deep structures give all the information required for determining the meaning of sentences. Thus, he now distinguishes between the initial phrase markers (the former deep structures), which initiate transformations, and the syntactically enriched surface structures, which undergo semantic interpretations.

The essence of Chomsky’s criticism of distributional grammar is that it does not provide an adequate theory of linguistic competence. In Chomsky’s view, the fact that speakers are able to produce and understand novel sentences demonstrates that they have a knowledge of the system of language, which explains those regularities in speech described in distributional grammars. The proper task of linguistics is to describe the underlying system, and to do that is to describe the set of syntactic, semantic, and phonological rules that generate (assign structural descriptions to) the grammatical sentences of some language.⁹ Chomsky’s hypothesis is that linguistic competence (that is, tacit knowledge of rules) underlies and accounts for a native speaker’s intuitions about the grammatical acceptability or unacceptability, the ambiguity or equivalence, and, in general, the meaning of any subset of an infinite set of possible grammatical sentences.

It follows from Chomsky’s view of competence that to acquire a language signifies constructing, in some manner, a representation of the set of rules that generate those utterances to which the language learner is exposed. Since the sample of utterances is necessarily finite, and since moreover it may contain ungrammatical, incomplete, or in some other sense degenerate data, the language learner must be engaged in a complex process of hypothesis testing. Furthermore, on Chomsky’s account, the structural relationships and ambiguities of which speakers are intuitively aware (and which must, therefore, be captured by explicit rules), can be fully described only by a transformational grammar, which posits a distinction between deep structure and surface structure. The competent speaker must, then, make inferences from the surface structures directly available in speech to more abstract phrase markers that may describe other kernel sentences. Thus, one cannot assume that language is acquired simply through exposure to the distributional structure implicit in speech. Distributional grammar is, again, said to be inadequate as a model of competence.

If Chomsky’s account of competence is correct, it is clear that language acquisition is an impossibly difficult task. Not only must the learner contend with noisy data, but she must recognize something about the nature of the task as a process of theory construction and hypothesis testing, and she must discover the crucial difference between underlying phrase markers and surface structures to construct an adequate grammar. Furthermore, of course, the language learner must do all of this at an age when there is no reason to believe human beings are capable of such sophisticated conceptual feats. In short, an adequate T-G grammar would seem to be, for children, and perhaps for most adults, unlearnable.

One might conclude from this argument that Chomsky’s view of competence must be wrong. But the theory of competence as knowledge of an adequate T-G grammar is justified by the argument that such a grammar accounts for creativity, since it enables a speaker to generate an infinite number of possible sentences. Thus, he says, if such a grammar cannot be learned in the ordinary sense (or in a way made precise by some learning theory), yet speakers must be assumed to have such knowledge, then there must be an innate basis for language acquisition. This suggestion is reinforced by the observation that language acquisition appears to be uniform across the species. All human beings acquire language—unless some physical defect prevents them from doing so—regardless of differential intelligence; limited, frequently degenerate, and widely variable speech data; and absence of formal instruction or apparent reinforcement. Moreover, the onset of speech activity is fairly constant, and, in Chomsky’s view, the acquisition period strikingly brief.¹⁰ All of these factors seem to point to a genetically based, specifically linguistic, cognitive ability that Chomsky characterizes as an innate, a priori knowledge of universal grammar.

Universal grammar (UG) is said to provide “a schema to which any particular grammar must conform (Chomsky, 1968, p. 76). It includes what Chomsky calls formal and substantive universals and an evaluation procedure or “weighing device” for selecting one grammar from among several possibilities. Formal univer...

Table of contents

• Acknowledgments
• Introduction
• Part I: Pragmatics and Semantics
• Part II: An Empiricist Theory of Utterance Meaning
• References
• Index