Chomsky has defined the goals of generative grammar as the explicit rendition of ''the ideal speaker-hearer's intrinsic competence.'' Other linguists sometimes use the phrases "what an adult native speaker knows" or "what every native speaker knows" or "What all native speakers know" These alternative expression for linguistic competence seem harmless enough, but in fact, they reveal a fundamental problem involved in the concept of linguistic competence and, hence, in the current linguistic enterprise as a whole. Let us see why this is so.
The alternative expressions are more grossly infelicitous than Chomsky's phraseology because if one were to take them at their face value, they would elude from the study of linguistics any but the most common and prosaic aspects of language. Just to focus on one obviously .absurd implication, it would exclude from its grammar any lexical item that isn't known by the whole population of speakers, which is to say the vast majority of all words in a literate language. The argument applies to the syntactic aspects of language, where linguistics s won't be reduced to the study of the lowest common denominator of inarticulate speakers. Chomsky's more careful wording avoids these absurdities and has the further merit of suggesting a concern with knowledge under ideal conditions unencumbered by the psychological restrictions that impair speech production and intuitive reflections (e.g., low verbal ability and limited experience).
But Chomsky's wording, no less than the alternative phrases, leaves uncharacterized the essential goals of linguistics: an "ideal speaker-hearer" is also a "communicator" (let us for the moment disregard all the other things he is too) and possesses intrinsic communicative competence. For example, he can tell whether his interlocutor is speaking seriously or in jest, he can use information he has about the interlocutor to interpret his utterances (e.g., the political party he belongs to and whether he had a domineering mother) and his knowledge of whether the interlocutor is a stranger, a friend, or a professional foe would undoubtedly affect the inferences he draws about what his sentences imply. In other words, since linguists must concern itself with the interpretation of sentences, what kind of restrictions does Chomsky propose to set with respect to the nature of the information for such interpretation that would properly fall within the confines of linguistic investigations? An answer to this question would require a fairly specific proposal for that component of language which Chomsky calls "semantic representation," and since he has not concerned himself (in his published writings) with that aspect as thoroughly as with the syntactic and phonological components, we must examine the well-known proposal of Katz and Fodor (1963) who have considered the very same issue that has been raised here and attempted a solution within the Chomskyan framework.
Katz and Fodor recognize that speakers in communicative settings (actual or ideal) may use non-linguistic information in the interpretation of sentences. However, they exclude this kind information as proper for linguistics on the grounds that such information is not now systematized, and further, could not be systematized in the foreseeable future. Since systematization is a necessary condition for a formalize linguistics, we should examine its consequences There are two alternatives possible in the delimitation of a field of study. One is to identify the phenomena to be concerned with on a priori grounds as to their significance and then apply to them whatever techniques of analysis are available at the time. The other is to commit oneself to a degree of formality available at the time and exclude from study any phenomenon to which the available formalized techniques are inapplicable.
The latter approach is more likely to lead to significant advances because increments in knowledge have a cumulative character, but this is true only in so far as the available formalized techniques are reasonably powerful. The S-R paradigm in behavioristic psychology went wrong on just this account, i.e., the formalization it offered was too weak so that all but the least significant of the phenomena had to be excluded from study. Transformational generative linguistics is a powerful formalization for syntactic phenomena and has enjoyed notable success in that area. It appears from the Katz and Fodor attempt the formalization that worked so well for syntax may be too weak or unsuitable for semantics.6 To appreciate these problems we must examine just how much the Katz and Fodor proposal is inadequate and to consider whether this inadequacy is specific to this instance or whether it is of a more general sort, endemic of the formalitization requirements of transformational generative linguistics.
The are at least two types of difficulties with the Katz and Fodor proposal. One is internal and has to do with the characteristics of the semantic features they postulate and the nature of the projection rules they propose The other is external and relates to the psychological processes that must be postulated to account for the interpretations of utterances. Both Bolinger (1965) and Weinreich (1966) have shown the internal inadequacy of the Katz and Fodor proposal, and in addition, Weinreich has offered some specific alternative suggestions within the transformational-generative formalization. Although his proposal is incomplete7 it indicates, at least, that the transformation rule device can be powerfully applied to solve such problems as metaphor and figurative speech, phenomena that completely eluded the frozen "atomistic" approach of Katz and Fodor.8
It is the second type of difficulty mentioned above that Weinreich's solution leaves as unanswered as the Katz and Fodor proposal. Linguists are generally committed to the proposition that their task in explicating semantic interpretation must e restricted to the universe of words and the combinatorial processes that words enter into when linked in syntactic structure But if it can be shown that the interpretation of sentences typically involves psychological processes that fall outside this limitation, then the maintenance of such a restriction is achieved only by an unacceptable reduction of the meaning of "semantic interpretation."9 Recently, several writers familiar with the transformational-generative linguistics paradigm have argued against such reductionism and for the incorporation of additional levels of analyses for semantic interpretation (see Harman, 1958; Heny, 1969; Searle, 1969). Heny, for instance, who is an "M.I.T. linguist," has stated that "linguistic theory should include a semantics based on relationships between linguistic entities and conditions in the world" on the grounds that it is very odd to claim to repre sent an ideal speaker's linguistic competence when the theory is unable to specify how such a person knows the truth value of a simple indicative sentence (e.g., "There are six men in this room"). Or, to take another example' involving an indexical item such as the deictic "This," how can one determine the anomaly of:
This is ripe.
when said of a piece of coal unless linguistics takes into account the properties of objects in the environment of the utterance?10
Searle (1969), and no doubt others as well, has pointed to an essential difference between rule-governed systems such as games (e.g., chess) and language; and that is that, although a chess move and a sentence can both be viewed as generated by a system of "constitutive" rules, one does not mean anything by a chess move. The utterance of a sentence represents a speech act, and a person so acting means something by it. Thus the meaning of a sentence cannot be understood in the absence of ascribing some intention to a person who might have uttered it.
In discussions of "meaning," philosophers tend to emphasize reference (including propositions) and intention at the expense of syntactic considerations, while linguists concentrate on the combinatorial mechanisms and underlying structures at the expense of reference and intention. Some extreme positions have thus resulted. Thus, the philosopher Grice (1957) has stated that the meaning of an utterance is nothing more than the intention the speaker has in uttering it. Katz and Fodor have stated that the meaning of a sentence is to be computed without any consideration to reference and intention. It seems to me that both positions lead to infelicitous results. Consider the following example.
Exchange A: John: I am sorry you failed the exam.
Tom: The hell you are.
John: I am sorry you failed the exam.
What is the meaning of John's statement? What kind of theory do we have to construct to account for Tom's ability to understand the meaning of the sentence uttered by John?11 Note that John claims to be sorry, when in fact he is glad, and Tom gets angry at this hypocrisy. Consider, now, the hypothetical exchange B under the same circumstances:
John: I am glad you failed the exam.
Tom: I know you do, you S.O.B.
By limiting himself to intention exclusively, Grice would be unable to account for the difference in Tom's reply under exchanges A and B since, by his definition, the two sentences uttered by John have the same meaning. By limiting itself to a non-contextual analysis, Katz and Fodor's theory would not only be incomplete, but would present a false account of Tom's interpretation of John's utterance. Clearly, what is needed to account for a speaker-hearer's knowledge of how to interpret sentences is a theory that deals explicitly with the fact that sentences are a sub-set of a larger category of objects, called utterances; the latter have functional properties whose characteristics are discoverable only by relating them to the larger communicative context that is described by psychological, sociological, and anthropological investigations.
Weinreich (1966) has argued that the claim by Katz and Fodor that a linguistic theory is to give a description of how speakers interpret sentences is too broad a claim for linguistics, since that is a claim about language performance. Strangely enough, Weinreich then proceeds to present what he terms a linguistic theory of meaning that ! purports to represent how speakers can interpret sentences. Perhaps Weinreich wished to emphasize the distinction between "how speakers interpret sentences" and how speakers can interpret sentences"echoing Chomsky's insistence that a generative grammar is not a model of the processes in the brain when speakers use language. At any rate, the distinction between competence and performance is not an absolute one, and Chomsky has been unable to present a principled argument for their distinction.
Many of the new proposals in generative semantics, especially those concerned with presuppositions, topic, focus, quantifiers, and pronouns, are enlarging the domain of competence to elements and inferential abilities that fall in the domain of performance by earlier standards (such as those of Chomsky, 1965, and even more so, Chomsky, 1957). The rate at which so-called competence factors "encroach" upon so-called performance factors is partly determined by the constraints imposed upon linguists by the nature of transformation rules that they are willing to accept at any one time as being "proper" to language. Let the linguists fight it out among themselves. To me it seems that if we are serious in studying meaning, or the speaker's ability to interpret sentences, we must set our sights to something like "communicative competence" and not be bound by current constraints in linguistic theory. Let the linguists catch up with us later.
A theory of communicative competence does not claim to account for how person X interprets (or will interpret) sentence Y in context Z at time t. One might hope that, given sufficient knowledge about people, such an account of people's performances might be possible. But such a hope might not be justified, in the same sense that physicists are unable to predict the path of a rock rolling down a hill. In fact, this kind of prediction does not seem to be necessary or even useful for an understanding of the principles involved in the facts of the situation. Similarly, it seems to me, that all that could be hoped for in a theory of communicative competence is an explication of the principles that can operate in the situation where speakers use and interpret sentences. It is for this reason that the data to be considered in the process of working out such a theory need not be of an experimental sort as commonly understood by psychologists.12 Any competent person engaged in the systematic study of communicative competence has available to him, through prior experience and observation, as well as through "Gedanken" exercises, a wealth of empirical data far richer than any to be observed in the psychological laboratory.13 Perhaps sometime in the distant future, when our theory is sufficiently well worked out to be able to account for the known facts of everyday communication, formal experiments might be of greater use.
6. Generative linguists sometimes state that transformation theory is "too powerful" in that it is applicable to practically any system of phenomena and it is their task to find out what kind of restrictions must be imposed on transformation theory to make it a theory of language as opposed to a theory of any system. Whether or not the claim is valid, the point I wish to make here is that the restrictions that are to be imposed on a transformation theory of syntax may weaken it too much for a theory of semantics, while a transformation theory of semantics mat be too powerful for a theory of syntax. In other words, the two domains may not lend themselves to an explication by the same formalization.
7. How regrettable it is that this brilliant thinker died an untimely death with his most vigorous years of scientific work unfulfilled.
8. The major conceptual difference between the proposal of Katz and Fodor and that of Wienreich lies in the fact that the former applies semantic manipulations after all syntactic manipulations have been applied, while in the latter case, they are deeply interpenetrated, the two aspects no longer clearly distinguishable. By using such concepts as rules of suppression, transfer, modalization, redistribution, causation--which are psychological in nature--Weinriech succeeds in showing how some of the traditional stylistic devices in rhetoric can be handled within the transformational paradigm. See also the related proposals of Bendix (in press).
9. This is precisely the kind of reductionism behaviorists are guilty of when defining mentalistic concepts in behavioral terms and against which Chomsky has so vigorously argued.
10. One might argue that this difficulty can be resolved by substituting other linguistic elements for the indexical item, in which case the anomaly of the sentence "The piece of coal that lies on the table is ripe" is handled in the usual way. However, this solution fails generally since the substitutability of indexical items is problematic- cf. "I am afraid of something, I don't know what", That really makes me mad!," etc.
11 . We shall assume the simpler case where Tom's interpretation is correct. In the case where it is not, the account will have to be more complex, but still of the same sort.
12. One or the unfortunate consequences of the competence-performance distinction formulate by Chomsky has been that psychologists have interpreted it literally as a distinction between knowledge of principles and observed performance The task they have defined for themselves restricts their interest to records of people's behavior in controlled, and artificial,, experiments, thus denying themselves the rich source of data available to linguists and philosophers This is a kind of neo-neo-behaviorism that matches the poverty of neo-behaviorism and behaviorism of earlier times.
13. I do not wish to minimize the usefulness of data not accessible by these means such as the observations that can be made on the development of communicative competence in children But note that in such cases, the traditional experimental paradigm does not hold either.
Back to the Titlepage