280    Limits of the Literal
repetition; and in that case, convention does not help explain what is basic to linguistic communication, though it may describe a usual, though contingent, feature.
I have a final reflection. I have not argued here, though I have elsewhere, that we cannot confidently ascribe beliefs and desires and intentions to a creature that cannot use language.14 Beliefs, desires, and intentions are a condition of language, but language is also ä condition for them. On the other hand, being able to attribute beliefs and desires to a creature is certainly a condition of sharing a convention with that creature; while, if I am right in what I have said in this paper, convention is not a condition of language. I suggest, then, that philosophers who make convention a necessary element in language have the matter backwards. The truth is rather that language is a condition for having conventions.
14 See Essay 11.
i
Bibliographical References
Allen, Woody, 'Condemned', New Yorker (21 November 1977), 59. Arendt, H., The Life of (he Mind. Harcourt, Brace, Jovanovich, New
York (1978). Austini, J. L., and Strawson, P. F., 'Symposium on "Truth"', in
Proceedings of the Aristotelian Society, Supplementary Volume 24
(1950). Austin, J. L., How to Do Things with Words. Harvard University
Press, Cambridge, Mass. (1962). Bain, A., English Composition and Rhetoric. D. Appleton, New York
(1867). Bar-Hillel, Y., 'Logical Syntax and Semantics', Language, 30 (1954),
230-7. Bär-Hillel, Y., 'Remarks on Carnap's Logical Syntax of Language',
in The Philosophy of Rudolf Carnap, ed. P. A. Schilpp. Open
Court, La Salle, Illinois (1963). Bar-Hillel,   Y.,   Language   and   Information.   Addison-Wesley,
Jerusalem (1964). Barfield, O., 'Poetic Diction and Legal Fiction', in The Importance
of Language, ed. M. Black. Prentice-Hall, Englewood Cliffs, New
Jersey (1962). Benacerraf, P., 'Mathematical Truth', Journal of Philosophy, 70
(1973), 661-79. Beth, E. W., 'Carnap's Views on the Advantages of Constructed
Systems Over Natural Languages in the Philosophy of Science', in
The Philosophy of Rudolf Carnap, ed. P. A. Schilpp. Open Court,
La Salle, Illinois (1963). Black, M., 'The Semantic Definition of Truth', Analysis, 8 (1948),
49-63.
264   Limits of the Literal
hidden message, like the critic who attempts to state the message, is then fundamentally confused. No such explanation or statement can be forthcoming because no such message exists.
Not, of course, that interpretation and elucidation of a metaphor are not in order. Many of us need help if we are to see what the author of a metaphor wanted us to see and what a more sensitive or educated reader grasps. The legitimate function of so-called paraphrase is to make the lazy or ignorant reader have a vision like that of the skilled critic. The critic is, so to speak, in benign competition with the metaphor maker. The critic tries to make his own art easier or more transparent in some respects than the original, but at the same time he tries to reproduce in others some of the effects the original had on him. In doing this the critic also, and perhaps by the best method at his command, calls attention to the beauty or aptness, the hidden power, of the metaphor itself.
\
Communication and Convention
Convention figures conspicuously in many of our activities, for example in playing tarot, in speaking, and in eating. In playing tarot, convention is essential, in eating it is not. In explaining what it is to play tarot we could not leave out of account the rules that define the game; in explaining what it is to eat no mention of rules or conventions needs to be made. What is the case with speech? Are conventions mere conveniences or social flourishes, or are they necessary to the existence of communication by language?
The question is delicate because it concerns not the truth of the claim that speech is convention-bound, but the importance and role of convention in speech. The issue may be put counterfactually: could there be communication by language without convention? According to David Lewis, it is a platitude—something only a philosopher would dream of denying—that there are conventions of language.'1 Certainly it would be absurd to deny that many conventions involve speech, such as saying 'Good morning' no matter what the weather is like; but this is not the sort of convention on which the existence of language depends. No doubt what Lewis has in mind is the idea that the connection between words and what they mean is conventional. And perhaps only a philosopher would deny this; but if so, the reason may be that only a philosopher would say it in the first place. What is obvious enough to be a platitude is that the use of a particular sound to refer to, or mean, what it does is arbitrary. But while what is conventional is in some sense arbitrary, what is arbitrary is not necessarily conventional.
In one respect we describe a language completely when we say
18
1 D. Lewis, 'Languages and Language', 7.
266    Limits of the Literal
what counts as a meaningful utterance and what each actual or potential utterance means. But such descriptions assume we already know what it is for an utterance to have a particular meaning. Light on this question—the traditional problem of meaning—requires us to connect the notion of meaning with beliefs, desires, intentions, and purposes in an illuminating way. It is mainly in making the connection, or connections, between linguistic meaning and human attitudes and acts described in non-linguistic terms that convention is asked to do its work. And here there are many different theories that have been proposed. I shall divide them into three kinds: first, there are theories that claim there is a convention connecting sentences in one or another grammatical mood (or containing an explicit performative phrase) with illocutionary intentions, or some broader purpose; second, there are theories that look to a conventional use for each sentence; and third, there are theories to the effect that there is a convention that ties individual words to an extension or intension. These are not competing theories. Depending on details, all combinations of these theories are possible. I discuss the three sorts of theory in the order just listed.
In an early, and influential, article Michael Dummett maintained that there is a convention that governs our use of declarative sentences.2 As he has put it more recently:
... the utterance of a [declarative] sentence does not need a particular context to give it a point ... The utterance of a sentence serves to assert something ... there is a general convention whereby the utterance of a sentence, except in special contexts, is understood as being carried out with the intention of uttering a true sentence.3
This is a complex, and perhaps not entirely transparent, dictum, but I interpret it as follows. There is a conventional connection between uttering a declarative sentence and using it to make an assertion (one is making an assertion except in special contexts); and there is a conceptual (and perhaps conventional) connection between making an assertion and the intention to say what is true. The plausibility of this interpretation is brought out, I think, by Dummett's most convincing argument. He begins by examining Tarski-style truth definitions. Dummett reminds us (following, though he probably did not know this, an earlier paper by Max Black4) that while Tarski
2 M. Dummett, 'Truth'.
3 M. Dummett, Frege: Philosophy of Language, 298.
4 M. Black, 'The Semantic Definition of Truth*.
Communication and Convention   267
showed how, in principle, to construct a truth definition for particular (formalized) languages, he did not, and indeed proved that no one could, define truth in general, at least using his method. Tarski was therefore not able to say what it was that made each definition of truth a definition of the same concept. Convention T, to which Tarski appealed for a criterion of the correctness of a truth definition, does not specify what truth in general is, but makes use of the intuitive grasp we have of the concept.
Dummett drew an analogy between truth and the concept of winning at a game. If we want to know what winning at a game is, we will not be satisfied by being told the definition of winning for each of several games; we want to know what makes the situation defined for each game a case of winning. Thinking of truth, the problem could be put this way: if we were exposed to speakers of a language we did not know, and were given a Tarski-style truth definition, how could we tell whether the definition applied to that language? A good question; but I do not believe it can be answered by attending to Dummett's proposed convention. For it seems to me that nothing in language corresponds in relevant ways to winning in a game. The point is important because if Dummett is right, to understand what it is in language that is like winning in a game is to make the crucial connection between meaning as described in a theory of truth and the use of language in contexts of communication.5
Winning in a game like chess has these characteristics: first, people who play usually want to win. Whether they want to win or not, it is a condition of playing that they represent themselves as wanting to win. This is not the same as pretending they want to win, or trying to get others to believe they want to win. But perhaps representing oneself as wanting to win does entail that one can be reproached if it is found that he does not want to win or isn't trying to win. Second, one can win only by making moves defined by the rules of the game, and winning is wholly defined by the rules. Finally, winning can be, and often is, an end in itself.6 As far as I can see, no linguistic  behaviour  has  this  combination  of features;  if so,
s It is not pertinent to my argument that Dummett does not believe a theory of truth can serve as a theory of meaning. The issue here is whether or not there is a convention of a certain sort governing our utterances of (declarative) sentences.
b The distinction between activities that can be ends in themselves, such as playing the flute, and those that serve some further end, such as building a house, comes, of course, from Aristotle: Nichomachean Ethics 1094a; Magna Moralia 1211b.
268    Limits of the Literal
Dummetťs analogy between  games and  language  is  radically defective.
Is speaking the truth, in the sense of intentionally uttering a sentence that happens to be true, like winning? It is in this respect, that what it is to speak the truth is what a theory of truth seeks to define. In so far, then, as the truth conditions of utterances are known to speakers and interpreters in advance, and agreed upon as a condition of communication, speaking the truth has one of the features of winning. (I'll question how far this is true later.) But it lacks the others, for people who utter a sentence do not usually want to speak true sentences. Sometimes they do, and very often they don't. Nor, in order to play the speech game, do they have to represent themselves as intending or wanting to speak the truth; there is no general presumption that someone who utters a declarative sentence wants or intends to speak the truth, nor that, if he does, he does it intentionally. Finally, speaking the truth, in the sense of uttering a true sentence, is never an end in itself.
Assertion, in contrast to speaking the truth, may seem a likelier candidate for linguistic counterpart of winning. Someone who makes an assertion represents himself as believing what he says, and perhaps as beingjustified in his belief. And since we want our beliefs to be true, it seems right to agree with Dummett that when someone makes an assertion, he represents himself as intending to say what is true. (This is how I take Dummetťs remark that the speaker is 'understood' to have the intention of uttering a true sentence.) As in playing a game, the representation may or may not be deceitful. (The liar makes an assertion.) The asserter may or may not, in making his assertion, intend to cause his hearer to believe he believes what he says. Making an assertion is, then, like playing a game in a respect in which speaking the truth is not: there is a public presumption of purpose. In other respects, however, assertion is unlike winning, for what constitutes the making of an assertion is not governed by agreed rules or conventions.7
If the concept of assertion is to provide a conventional bridge between purpose and truth, two things must hold: there must be conventions governing assertion, and there must be a convention linking assertion to what is believed true. I think neither of these claims holds.
7 On this point I am much indebted to Sue Larson.
Communication and Convention    269
Many philosophers have thought there were conventions governing assertion. Thus Dummett, in a phrase I omitted from an earlier quotation, says, 'The utterance of a sentence serves to assert something .. .'8 Let us first consider whether assertion is governed by conventions. Of course, if it is a convention that a sentence means what it literally does when uttered, then convention is involved in all utterances and hence in assertion. But literal meaning may not (and in my view does not) go beyond truth conditions. And no one will deny, I suppose, that the same declarative sentence may have the same meaning when used to make an assertion, to tell a joke, to annoy a bore, to complete a rhyme, or to ask a question. So if there is a convention, it must be further conventional trappings of the utterance that make it an assertion. It is not enough, of course, to say that something in the context makes it an assertion. This is true, but proves nothing about convention. And we may be able to say what it is in the context that makes it an assertion, though in fact I think we can say only some rather vague and incomplete things. But even if the necessary and sufficient conditions were explicit and agreed upon by all hands it would not yet follow that the conditions were conventional. We all agree that a horse must have four legs, but it is not a convention that horses have four legs.
There is something more about assertion that suggests that convention may be involved, and this is the fact that in making an assertion, the asserter must intend to make an assertion, and he must intend that this intention be recognized by his audience. Assertions are intended to be public performances where the clues are adequate to identify the character of the performance as assertoric. So it is natural to think it would be useful if there were a convention, as a convenience in making our assertive intentions clear.
But Frege was surely right when he said, 'There is no word or sign in language whose function is simply to assert something.' Frege, as we know, set out to rectify matters by inventing such a sign, the turnstile V. And here Frege was operating on the basis of a sound principle: if there is a conventional feature of language, it can be made manifest in the symbolism. However, before Frege invented the assertion sign he ought to have asked himself why no such sign existed before. Imagine this: the actor is acting a scene in which there is supposed to be a fire. (Albee's Tiny Alice, for example.) It is
See footnote 3.
270   Limits of the Literal
his role to imitate as persuasively as he can a man who is trying to warn others of a fire. 'Fire!' he screams. And perhaps he adds, at the behest of the author, 'I mean it! Look at the smoke!' etc. And now a real fire breaks out, and the actor tries vainly to warn the real audience. 'Fireľ he screams, 'I mean it! Look at the smoke!' etc. If only he had Frege's assertion sign.
It should be obvious that the assertion sign would do no good, for the actor would have used it in the first place, when he was only acting. Similar reasoning should convince us that it is no help to say that the stage, or the proscenium arch, creates a conventional setting which negates the convention of assertion. For if that were so, the acting convention could be put into symbols also; and of course no actor or director would use it. The plight of the actor is always with 'us. There is no known, agreed upon, publicly recognizable convention for making assertions. Or, for that matter, giving orders, asking questions, or making promises. These are all things we do, often successfully, and our success depends in part on our having made public our intention to do them. But it was not thanks to a convention that we succeeded.
The second point of Dummett's claim is that there is a convention that in making an assertion a speaker is 'understood' to be speaking with 'the intention of uttering a true sentence'. This also seems to me to be wrong, though in, a somewhat different way. What is understood is that the speaker, if he has asserted something, has represented himself as believing it—as uttering a sentence he believes true, then. But this is not a convention, it is merely part of the analysis of what assertion is. To assert is,, among other things, to represent oneself as believing what one asserts. It is clear that there cannot be a conventional sign that shows that one is saying what one believes; for every liar would use it. Convention cannot connect what may always be secret—the intention to say what is true—with what must be public—making an assertion. There is no convention of sincerity.                    „
If literal meaning is conventional, then the difference in the grammatical moods—declarative, imperative, interrogative, optative—is conventional. These differences are in the open and intended to be recognized; syntax alone usually does the job. What this shows is that grammatical mood and illocutionary force, no matter how closely related, cannot be related simply by convention.
Although I have concentrated on assertion, similar considerations
Communication and Convention    271
apply to illocutionary forces of all kinds. My main interest here, however, is not in the nature of illocutionary force, or in such acts as asserting, promising, and commanding, but in the idea that convention can link what our words mean—their literal semantic properties, including truth—and our purposes in using them, for example, to speak the truth.
We have been discussing claims that there are comprehensive purposes tied by convention to the enterprise of linguistic communication—purposes that, in Dummett's word, give us the 'point' of using language. 1 turn now to theories of quite a different sort, that attempt to derive the literal meanings of entire sentences (not just the mood indicators) from the non-linguistic purposes their utterances serve. I am concerned in the present essay with theories which make the derivation depend on convention.
Stated crudely, such theories maintain that there is a single use (or some finite number of uses) to which a given sentence is tied, and this use gives the meaning of the sentence. Since in fact there are endless uses to which a sentence, with meaning unchanged, can be put, the connection between a single use (or finite number of uses) and the sentence is conventional; it is a use that can be called standard.
This is too simple, of course, but it is án appealing and natural idea. For there does seem to be an important connection between a sentence like 'Eat your eggplant* and the intension, in uttering this sentence, to get someone to eat his eggplant. Getting someone to eat his eggplant is, you might say, what the English sentence 'Eat your eggplant' was made to do. If this intuition could be explicitly stated and defended in a non-question-begging way, there would be promise of an account of literal meaning in terms of the ordinary non-linguistic purposes that always lie behind the utterances of sentences.
There are intentions embedded in all linguistic utterances such that if we could detect them we would usually know what the words uttered literally meant. For someone cannot utter the sentence 'Eat the eggplant' with the words literally meaning that someone is to eat his eggplant unless he intends the sentence to have that meaning, and intends his audience to interpret it as having that meaning. Of course the mere intention does not give the sentence that meaning; but if it is uttered with the intention of uttering a sentence with that meaning, and it does not in fact have that meaning, then it has no linguistic meaning at all. Literal meaning and intended literal
272   Limits of the Literal
meaning must coincide if there is to be a literal meaning. But this fact, while true and important, is of no direct help in understanding the concept of literal meaning, since the crucial intention must be characterized by reference to the literal meaning. Nor can convention make a contribution here, for we were looking to convention to convert non-linguistic purposes into performances with a literal meaning. A convention that connected the intention to use words with a certain literal meaning with the literal meaning of those words would not explain the concept of literal meaning, but would depend upon it.
What we seek are intentions characterized in non-linguistic terms—ulterior purposes in uttering sentences. (This concept may be related to what Austin called perlocutionary acts.)
I mentioned briefly before, and now want to insist on, the.fact that linguistic utterances always have an ulterior purpose; this was one of the reasons I gave for claiming that no purely linguistic activity is like winning at a game. ^There is perhaps some element of stipulation here, but I would not call it a linguistic act if one spoke 'words' merely to hear the sounds, or to put someone to sleep; an action counts as linguistic only if literal meaning is relevant. But where meaning is relevant, there is always an ulterior purpose. When one speaks, one aims to instruct, impress, amuse, insult, persuade, warn, remind, or aid a calculation. One may even speak with the intention of boring an audience; but not by hoping no one will attend to the meaning.
If I am, right that each use of language has an ulterior purpose, then one must always intend to produce some non-linguistic effect through having one's words interpreted. Max Black has denied this, pointing out that '. .. a man may outline a lecture, or write a note to remind himself of an appointment, or simply utter certain words, such as "What a lovely day!" in the absence of an audience.'9 The first two cases here are clearly cases^ where the meaning matters, and there is an audience which is intended to interpret the words: oneself at a later time. In the last case it would be tendentious to insist that one is speaking to oneself; yet it matters what words are used, what they rnean. And there must be some reason for using those words, with their meaning, rather than others. Black quotes Chomsky to similar effect:
9 M. Black, 'Meaning and Intention: An Examination of Grice's Views', 264.
Communication and Convention    273
Though consideration of intended effects avoids some problems, it will at best provide an analysis of successful communication, but not of meaning or the use of language, which need not involve communication, or even the attempt to communicate. If I use language to express or clarify my thoughts, or with the intent to deceive, to avoid an embarrassing silence, or in a dozen other ways, my words have a strict meaning and I can very well mean what I say, but the fullest understanding of what I intend my audience (if any) .to believe or do might give little or no indication of the meaning of my discourse.10
In this passage it seems to me Chomsky arrives at a correct conclusion from confused or irrelevant premises. The issue is whether or not the meanings of sentences can be derived from the non-linguistic intentions of a speaker. Chomsky concludes, correctly, J believe, that they cannot. But it is irrelevant to the conclusion whether intended effects must involve someone other than the speaker, and unimportant to the argument whether there are intended effects. Speaking or writing in order to clarify thoughts certainly posits an intended effect. Nor does it matter just how we use the word 'communicate'. What matters is whether an activity is interestingly considered linguistic when meanings are not intended to be put to use. Lying is a case where literal meaning is essential; the liar has an ulterior purpose t that is served only if his words are understood as having the meaning he intends.
Where Chomsky is right, as I said, is in claiming that no amount of knowledge of what I intend my audience to believe or do will necessarily yield the literal meaning of my utterance. Even this claim must, as we have seen, be limited to a description of my intention in non-linguistic terms. For if I intend to get my audience to do or believe something, it must be through their correct interpretation of the literal meaning of my words.
It is now relatively clear what a, convention must do if it isto reíate non-linguistic purposes in uttering sentences—ulterior purposes— with the literal meanings of those sentences when uttered. The convention must pick out, in a way understood by both speaker and hearer, and in an intentionally identifiable way, those cases in which the ulterior purpose directly yields the literal meaning. I mean, for example, a case where, in uttering the words 'Eat'your eggplant' with their normal meaning in English, a speaker intends to get a hearer to eat his eggplant through the hearer's understanding of the
10 N. Chomsky, Problems of Knowledge and Freedom, 19.
274     Limits of the Literal
words and the illocutionary force of the utterance. And here, once again, it seems to me not only that there is no such convention, but that there cannot be. For even if, contrary to what I have argued, some convention governs the illocutionary force of the utterance, the connection with the intention that the request or order be carried out would require that the speaker be sincere—that what he represents himself as wanting or trying to do he in fact wants or is trying To do. But nothing is more obvious than that there cannot be a, convention that signals sincerity.
It is no help, I must repeat, to say that the convention is that the
sentence always means what the ulterior purpose would reveal if the
speaker were sincere, serious,' etc. This is at best a partial analysis of
the relations between literal meaning, sincerity, and intention. It
' does not suggest publicly recognized tests, criteria, or practices.
Sometimes it is suggested that a language could never be learned except in an atmosphere of-hqnest assertions (commands, promises, etc.). Even if this were true, it would prove nothing about a supposed role for convention. But I am also sceptical about the claim itself, partly because so much language learning takes place during games, in hearing stories, and in pretence, and partly because the (acquisition of language cannot to such an extent depend on our luck in having truthful, sober, assertive playmates and parents.
It is in the nature of a game ljke chess or tarot not only that there are mutually agreed criteria of what it is to play, but mutually agreed criteria of what it is to win. It'is essential to these games that normally there is no question about the outcome. And it is also in the nature of such games that winning can be an end in itself, and that players represent themselves as wanting or trying to win. But the criteria for deciding what an utterance literally means, given by a theory of truth or meaning for the speaker, do not decide v/hether he has'accomplished his- ulterior purpose, nor is there any general rule that speakers represent themselves as having any further end than that of using words with a certain meaning and force. The ulterior purpose may or may not be evident, and it rhay or may not help an interpreter determine the literal meaning. I conclude that it is not an accidental feature of language that the ulterior purpose of an utteranceand its literal meaning are independent, in the jsense Jhat the latter cannot be derived from the foim£ĽJULÍs_Qri3iie essence of language. I call this feature of language the principle of the autonomy öf'meäning.  We came  across  an  application  when  discussing
I
/
Communication and Convention   275
illocutionary force, where it took the form of the discovery that what is put into the literal meaning then becomes available for any ulterior (non-linguistic) purpose—and even any illocutionary performance.'l
Before leaving the discussion of theories of the first two kinds, the following remark may be useful. Nothing I have said has been intended  to show there is  no connection  between  the. mood indicators and the idea of a certain illocutionary act. I believe there is such a connection. An utterance of an imperative sentence, for example, quite literally labels itself as an act of ordering. But this is just part of the literal meaning of the uttered words, and establishes no relation, conventional or otherwise, between the illocutionary intentions of the speaker and his words. It is easy to confuse two quite different theses: on the one hand, the (correct) thesis that every utterance of an imperative labels itself (truly or falsely) an order, and the thesis that there is a convention that under 'standard' conditions the utterance of an imperative is an order. The first thesis does, while the second thesis can't, explain the difference in meaning between an imperative and a declarative sentence, a difference which exists quite independently of illocutionary force. The second thesis can't explain this because^ it postulates a convention that is in force only under 'standard' conditions. You can't use a convention by breaking it, you can only abuse it. But the difference between an imperative and a declarative can, and very often is, quite properly used in situations where mood and illocutionary force are not 'standard'. It will' not help to point to such cases as an actor wearing a crown to indicate that he is playing the part of a king. If a convention is involved here, it is a convention governing literal meaning. Wearing a crown, whether done in jest or in earnest, is just like saying 'I am the king'. The gist of this remark applies also to the kind of theory that tries to derive the literal meaning of each sentence from a 'standard' use. Since the literal meaning operates as well when the use is absent as when it is present, no convention that operates only in 'standard' situations can give the literal meaning.
We have considered the idea that linguistic activity in general is like a game in that there is a conventional purpose (saying what is true, winning) which can be achieved only by following, or using, agreed and public rules. Then we discussed the claim that the literal
11 For further discussion, and a suggestion as to how illocutionary force and
grammatical mood are related, see Essay 8.
I
276    Limits of the Literal
meaning of each sentence is related by a convention to a standard non-linguistic end (an ulterior purpose). Both of these views turned out, on examination, to be untenable. Now it is time to evaluate the 'platitude' that the meaning of a word is conventional, that is, that it is a convention that we assign the meaning we do to individual words and sentences when they are uttered or written.
According to David Lewis12 a convention is a regularity R in action, or action and belief, a regularity in which more than one person must be involved. The regularity has these properties:
(I) Everyone involved conforms to R and (2) believes that others also conform. (3) The belief that others conform to R gives all involved a good reason to conform to R. (4) All concerned prefer that there should be conformity to R. (5) R is not the only possible regularity meeting the last two conditions. (6) Finally, everyone involved knows (l)-(5) and knows that everyone else knows (l)-(5), etc.
Tyler Bürge has raised reasonable doubts about parts of the last condition (does a convention require that everyone know there are alternatives?)13 and I have misgivings myself. But here there is no reason to debate the details of the analysis of the concept of convention; what is relevant is to raise a question whether, or to what extent, convention in any reasonable sense helps us understand linguistic communication. So instead of asking, for example, what 'conforming' to a regularity adds to the regularity itself, I shall simply grant thatisomething like Lewis's six conditions does hold roughly for what we call speakers of the same language. How fundamental a fact is this about language?
Lewis's analysis clearly requires that there be at least two people involved, since convention depends on a mutually understood practice. But nothing in the analysis requires more than two people. Two people could have conventions, and could share a language.
What exactly is t'he necessary convention? It cannot be that speaker and hearer mean the same thing by uttering the same sentences. For such conformity, while perhaps fairly common, is not necessary to communication. Each speaker may speak his different language, and this will not hinder communication as long as each hearer understands the one who speaks. It could even happen that
12  D. Lewis. 'Languages and Language", 5. 6.
13 T. Bürge. 'Reasoning about Reasoning*.
Communication and Convention    277
every speaker from the start had his own quite unique way of speaking. Something approaching this is in fact the case, of course. Different speakers have different stocks of proper names, different vocabularies, and attach somewhat different meanings to words. In some cases this reduces the level of mutual understanding; but not necessarily, for as interpreters we are very good at arriving at a correct interpretation of words we have not heard before, or of words we have not heard before with meanings a speaker is giving them.
Communication does not demand, then, that speaker and hearer mean the same thing by the same words; yet convention requires conformity on the part of at least two people. However, there remains a further form of agreement that is necessary: if communication succeeds, speaker and hearer must assign the same meaning to the speaker's words. Further, as we have seen, the speaker must intend the hearer to interpret his words in the way the speaker intends, and he must have adequate reason to believe that the hearer will succeed in interpreting him as he intends. Both speaker and hearer must believe the speaker speaks with this intention, and so forth; in short, many of Lewis's conditions would seem to be satisfied. It's true that this is at best an attenuated sense of a practice or convention, remote from the usual idea of a common practice. Still, one might insist that this much of a mutually understood method of interpretation is the essential conventional core in linguistic communication.                    .
But the most important feature of Lewis's analysis of convention—regularity—has yet to be accounted for. Regularity in this context must mean regularity over time, not mere agreement at a moment. If there is to be a convention in Lewis's sense (or in any sense, I would say), then something must be seen to repeat or recur over time. The ' only candidate for recurrence we have is the interpretation of sound patterns: speaker and hearer must repeatedly, intentionally, and with mutual agreement, interpret relevantly similar sound patterns of the speaker in the same way (or ways related by rules that can be made explicit in advance).
I do not doubt that all human linguistic communication does show a degree of such regularity, and perhaps some will feel inclined to make it a condition of calling an activity linguistic that there should be such regularity. I have doubts, however, both about the clarity of the claim and its importance in explaining and describing
278    Limits of the Literal
communication. The clarity comes into question because it is very difficult to say exactly how speaker's and hearer's theories for interpreting the speaker's words must coincide. They must, of course, coincide after an utterance has been made, or communication is impaired. But unless they coincide in advance, the concepts of regularity and convention have no definite purchase. Yet agreement on what a speaker means by what he says can surely be achieved even though speaker and hearer have different advance theories as to how to interpret the speaker. The reason this can be is that the speaker may well provide adequate clues, in what he says, and how and where-he says it, to allow a hearer to arrive at a correct interpretation. Of course the speaker must have some idea how the hearer is apt to make use of the relevant clues; and the hearer must know a great deal about what to expect. But such general knowledge > i is hard to reduce to rules, much less conventions or practices.
It is easy to misconceive the role of society in language. Language is, to be sure, a social art. But it is an error to suppose we have seen deeply into the heart of linguistic communication when we have noticed how society bends linguistic habits to a public norm. What is conventional about language, if anything is, is that people tend to speak much as their neighbours do. But in indicating this element of the conventional, or of the conditioning process that makes speakers rough, linguistic facsimiles of their friends and parents, we explain no more than the convergence; we throw no light on the essential nature of the skills that are thus made to converge. ,    This  is, not  to  deny  the  practical,  as contrasted  with  the theoretical,  importance of social  conditioning.  What i common conditioning ensures is that we may, up to a point, assume that the same method 'of interpretation that we use for others, or that we assume others use for us, will work for a new speaker. We do not have the time, patience, or opportunity to evolve a new theory of interpretation fcfo each speaker, and what saves us is that from the f moment someone unknown to us opens his mouth, we know an enormous amount about the sort of theory that will work for him— or we know we know no such theory. But if his first words are, as we say, English, we are justified in assuming he has been exposed to linguistic conditioning similar to ours (we may even guess or know differences). To buy a pipe, order a meal, or direct a taxi driver, we go on this assumption. Until proven wrong; at which point we can revise our theory of what he means on the spur of the moment. The
'
Communication and Convention   279
longer talk continues the better our theory becomes, and the more finely adapted to the individual speaker. Knowledge of the conventions of language is thus a practical crutch to interpretation, a crtftch we cannot in practice afford to do without—but a crutch which, under optimum conditions for communication, we can in the end throw away, and could.in theory have done without from the start.
The fact that radical interpretation is so commonplace—the fact, that is, that we use our standard method of interpretation only as a useful starting point in understanding a speaker—is hidden from us by many things, foremost among them being that syntax is so much more social than semantics. The reason for this, roughly stated, is that what forms the skeleton of what we call a language is the pattern of inference and structure created by the logical constants: J- the sentential connectives, quantifiers, and devices for cross-reference. If we can apply our general method of interpretation to a speaker at all—if we can make even a start in understanding him on the assumption that his language is like oure, it will thus be because we can treat his structure-forming devices as we treat ours. This fixes the logical form of his sentences, and determines the parts of speech. No doubt some stock of important predicates must translate in the obvious homophonic way if we are to get far fast; but we can then do very well in interpreting, or reinterpreting new, or apparently familiar, further predicates.
This picture of how interpretation takes place puts the application of formal methods to natural language in a new light. It helps show why formal methods are at their best applied to syntax; here at least there is good reason to expect the same model to fit a number of speakers fairly well. And there is no clear reason why each hypothesized method of interpretation should not be a formal ' semantics for what we may in a loose sense call a language. What we cannot expect, however, is that we can formalize the considerations. • that lead us to adjust our theory to fit the inflow of new information. No doubt we normally count the ability to shift ground appropriately as part of what we call 'knowing the language'. But in. this sense, there is no saying what someone must know who knows the language; for intuition, luck, and skill must play as essential a role here as in devising a new theory in any field; and taste and sympathy a larger role.
In conclusion, then, I want to urge that linguistic communication does not require, though it very often makes use of, rule-governed