Applications

Distinguishing between the three levels of meaning can clarify many issues in philosophy and linguistics. I shall briefly give some examples.

1 The distinction of levels tends to dissolve as verbal certain philosophical worries about what has to be true before someone can be said to use a language. One may use a system of representation in thinking, without being able to use it in communication or speech acts. Children and animals presumably do so, and perhaps some computers may also be said to do so. Similarly one may use a system of representation in thought and communication without being able to engage in more sophisticated speech acts. (Compare computers that ‘communicate’ with the programmer.) Whether communication or more sophisticated speech acts must be possible before one’s system of representation counts as a language can only be a purely verbal issue.

A special case of this issue would be the philosophical question whether there can be a private language. For the issue is simply whether there could be a language used to think in but not to communicate with. There can be a system of representation with such properties; whether it should count as a language is a purely verbal issue. On the other hand, Wittgenstein’s private-language argument may be directed against a conception of language learning and of the use of language in communication similar to that put forward in transformational linguistics by Chomsky, Katz, and Fodor, among others. I shall argue below that this conception is based on failure to distinguish levels of meaning.

2 The distinction can be used to help clarify various philosophical accounts of meaning. For example in Word and Object Quine presents considerations mainly relevant to level 1 theories of meaning. But by describing language as a set of dispositions to verbal behavior, he suggest wrongly that he is concerned with communication or more sophisticated speech acts. And this occasionally leads him wrong. He describes the thesis of indeterminacy as the view that a speaker’s sentences might be mapped onto themselves in various ways without affecting his dispositions to ‘ verbal behavior ’. So stated the thesis would be obviously wrong. A conversation containing one sentence would be mapped onto one containing another. Dispositions to verbal behavior would therefore change under the mappings in question. Actually Quine is interested in only one particular sort of verbal behavior: assent or dissent to a sentence. And his position would be even clearer if he had entirely avoided the behavioristic formulation and spoken instead about a speaker’s accepting as true (or accepting as false) various sentences.¹

In his papers on meaning, Paul Grice presents a level 2 theory of meaning. But in a recent paper² he is troubled by (among other things) (a) difficulties he has in accounting for the difference between telling someone one wishes him to do something and ordering him to do it, and (b) difficulties in accounting for meaning something by one’s words in silent thought. But (a) can be handled only within a level 3 theory, and (b) can be handled only within a level 1 theory. The former point is somewhat obscured by Grice’s formulation of the notion to be analyzed: ‘U meant x by uttering y.’ The locution is at least three ways ambiguous. It may mean (i) that x is the message conveyed by U’s uttering y, (ii) that U intended to say x when he said y, or (iii) that U really meant it when he said x; i.e., he uttered y with no fingers crossed, not ironically, not in jest, etc. Grice does not make clear exactly which of these interpretations we are to assign to the locution he is analyzing. A theory of communication results if the interpretation is (i). If (ii) were the correct interpretation, Grice’s analysis of meaning in terms of the speaker’s intentions would be trivialized. And (iii) involves appeal to speech acts. I urge Grice to accept (i), but I fear that he may opt for (iii). That would blur the line between levels 2 and 3 and would, I think, make it much more difficult for Grice to succeed in giving any sort of general account of meaning.

Alston presents a level 3 theory of meaning. But he believes that such a theory must account for sameness of meaning of linguistic expressions. I have argued above that this cannot be done. We cannot define sameness of meaning of expressions as sameness of illocutionary-act potential. Sameness of meaning is to be accounted for, if at all, within a level 1 theory. Given a theory of level 1, we might hope to define sameness of meaning (i.e., significance) of illocutionary acts via sameness of meaning of linguistic expressions. None of this shows that meaning cannot be approached via speech acts, as long as it is understood what sort of theory of meaning a theory of speech acts is.

3 The distinction between levels of meaning can be used to show what is really wrong with Katz and Fodor’s³ semantic theory. They claim that an adequate semantic theory must show how the meaning of a sentence is determined by its grammatical structure and the meaning of its lexical items. They say that such a theory must specify the form of dictionary entries for lexical items and must say how such entries are combined, on the basis of grammatical structure, in order to give readings of sentences.

These claims are the direct result of failure to distinguish a theory of the meaning of language as it is used in thinking from a theory of the meaning of a message, plus a failure to remember that in the standard case one communicates with a language one thinks in. Thus at first Katz and Fodor purport to be describing the structure of a theory of linguistic communication. They are impressed by the fact that a speaker has the ability to produce and understand sentences he has never previously encountered. As a result they treat communication as involving a complex process of coding and decoding, where readings are assigned to sentences on the basis of grammatical structure and dictionary entries. That this is a mistake has already been noted above. In normal linguistic communication a message is interpreted as expressing the thought (or some simple function of that thought) that is expressed by the same words the message is in.

The fact that a speaker can produce and understand novel sentences is a direct consequence of the facts that he can think novel thoughts and that he thinks in the same language he communicates in. One gives an account of the meaning of words as they are used in thinking by giving an account of their use in the evidence-inference-action game. For a speaker to understand certain words, phrases, and sentences of his language is for him to be able to use them in thinking, etc. It is not at all a matter of his assigning readings to the words, for to assign a reading to an expression is simply to correlate words with words.

Katz and Fodor do have some sense of the distinction between levels i and 2. Although their theory is put forward as if it were an account of communication, they describe it as a theory of meanings a sentence has when taken in isolation from its possible settings in linguistic discourse. They do this in order to avoid having to take into account special ‘ readings ’ due to codes, figurative uses of language, etc. Their theory of meaning is restricted to giving an account of the meaning of a message for that case in which the message communicates the thought that is expressed in the same words as those in which the message is expressed. They recognize that another theory would have to account for the interpretation that is assigned when a sentence occurs in a particular context.

In a way this amounts to distinguishing my levels 1 and 2. And in a sense Katz and Fodor attempt to provide a theory of level 1. More accurately, their theory falls between levels 1 and 2. It cannot provide a level 1 theory, since a speaker does not understand the words he uses in thinking by assigning readings to them. It cannot provide a level 2 theory, since it treats a very simple problem of interpretation as if it were quite complicated.

I think that perfectly analogous complaints can be raised against Paul Ziff’s⁴ views about meaning and against theories like Davidson’s⁵ that attempt to account for meaning in terms of truth conditions.

4 The foregoing points may shed some light on some of the puzzling things Chomsky says about linguistic competence and language learning.⁶ He says that anyone who knows a language has (unconscious) knowledge of the grammatical rules of the language. This is puzzling, since we would ordinarily ascribe knowledge of the rules of grammar (unconscious or conscious) to a linguist or grammarian. We would not ordinarily ascribe such knowledge to a typical speaker of the language. What is more puzzling is that Chomsky does not argue for this claim but takes it to be obvious. In the same vein, Chomsky treats language learning as a special case of theory construction. The child learning a language must infer a theory of the language that is spoken by those around him, i.e., he must infer a grammar of that language. Again this is puzzling because it treats the child as if he were a linguist investigating some hitherto untranslated language.

It is easy to point out the counterintuitive nature of Chomsky’s proposals and the difficulties involved if one takes them seriously. It is less easy to say what led Chomsky to make such proposals and why he continues to accept them in the face of heavy criticism. But seeing what is wrong with Katz and Fodor’s theory of semantics suggests what may be wrong with Chomsky’s views on linguistic competence and language learning.

To the extent that one forgets that people think in language as well as communicate in it, one will treat understanding a sentence as if it involved a complex process of decoding that uses the grammatical rules of the language in order to arrive at a reading or semantic interpretation of a sentence. One will take the grammar of a language to be, as it were, a code book, used by speaker and hearer in communication, since the grammar specifies what Chomsky calls ‘ the sound-meaning connections’.⁷ Therefore one will take it to be obvious that the speaker and hearer have (unconscious) knowledge of the rules of grammar. One will also make the mistake of treating language learning as simply a matter of learning how to communicate thoughts to others and how to understand others when they communicate. One will forget that in learning a language a person acquires a new system of representation that he can use in thinking. One will forget the relatively simple procedure we have for interpreting what others say to us, and one will be struck with the fantastic complexity of the code book, which (one thinks) the child acquires in an incredibly short time. One will describe this as theory construction. Impressed by the way in which the child’s theory is underdetermined by his evidence one will be led into rationalist speculation about innate ideas. And so on.

I suggest that this provides the most plausible account of what mistake has led Chomsky to say the strange things he has said about linguistic competence and language learning.

In this paper I have distinguished three levels in the theory of meaning corresponding to the meaning of thoughts, the meaning of messages, and the meaning of speech acts. I have argued that distinguishing these levels helps to clarify three well-known approaches to the theory of meaning and reveals certain deficiencies in Katz and Fodor’s semantics and in Chomsky’s discussion of linguistic competence and of language learning.

¹ Cf. my ‘ An Introduction to Translation and Meaning’: Word and Object, Synthese, XIX, 1-2 (December 1968): 14-26.

² ‘ Utterer’s Meaning and Intentions’, Philosophical Review, LXXVIII, 2 (April 1969): 147-77.

³ Katz and Fodor, ‘The Structure of a Semantic Theory’, in Fodor and Katz, eds., The Structure of Language.

⁴ Paul Ziff, Semantic Analysis (Ithaca, N.Y.: Cornell, 1960).

⁵ Donald Davidson, ‘Truth and Meaning’, Synthese, XVII, 3 (September 1967): 304-23.

⁶ E.g., in Aspects of the Theory of Syntax (New York: Harper & Row, 1965). Cf. my ‘ Psychological Aspects’ cited in n. a, p. 70, above and the discussion in the NYU Institute of Philosophy Proceedings for 1968, in Sydney Hook, ed., Language and Philosophy (New York: New York University Press, 1969).

⁷ E.g., in ‘ The Formal Nature of Language,’ Appendix A to Eric H. Lenneberg, Biological Foundations of Language (New York: Wiley, 1967), p. 397.