INTRODUCTION 4 страница

acceptable will, of course, depend on the adoption of particular

assumptions.

It is not difficult to find or construct similar examples in which, compound clauses containing but can be used after the verb 'say' in what appears to be, at least, the meaning of'assert' (in the logician's sense of'assert'). This does not prove that but contributes something other than what and contributes to the propositional content of such clauses. What it does show, however, is that the distinction between what is said and what is conventionally implicated is not always clear in the everyday use of the verb 'say'. More important, it also shows how the lexical and grammatical resources of a particular language can be adapted and exploited to propositionalize what is not of its nature propositional. This point is of the greatest importance. It will be

taken up and given further exemplification in Chapter 10 with particular reference to modality and subjectivity.

The only other example that Grice himself gave in his 1967/8 lectures to illustrate his notion of conventional implicature is the use of therefore. Once we look at the full range of languageuse, however, rather than simply at more or less formal argumentation, as Grice does, we can extend the list of forms which meet his criteria for conventional implicature very considerably.

Many of the connectives that serve to give cohesion to a text, linking one text unit with another, fall within the scope of his definition: however, moreover, nevertheless and yet, etc. So too do

 

9.4 Implication and conventional implicatures 275

modal particles such as even, well, or just, as in the following utter­ances:

Even Horace likes caviare,

You may well be right,

(20 It was just one of those things.

English, like French, has relatively few modal particles, in com­parison with German, Russian and many other languages. But, as examples (18)-(20) demonstrate, it does have some. More­over, their meaningfulness and their conventionality is evident from the fact that they can be mistranslated; and it is worth noting that mistranslation is possible even where exact transla­tion is not.

A second point to be made is that there seems to be no reason to restrict the notion of conventional implicature to connectives and particles. As we saw in Part 2, many fully lexical expressions are descriptively synonymous, but differ in respect of their social and expressive meaning. Most, if not all, of this difference would seem to fall within the scope of Grice's definition of con­ventional implicature. That is to say, morphological and syntac­tic distinctions, as well as differences between lexemes and particles, may be associated with what many semanticists, fol­lowing Grice, would classify as conventional implicatures. So, too, does much of the difference that is carried in particular contexts by the choice of one form of an expression, rather than another. For example, if the speaker says

(21) Christ tells us to love our neighbour

 

or

(22) Christ has told us to love our neighbour

rather than

(23) Christ told us to love our neighbour,

 

he or she can be held to have implicated that Christ's injunction or exhortation had, and still retains, a certain authority and validity. In fact, differences of tense and mood, not only in

 

276 Text and discourse; context and co-text

English but in many languages, are commonly associated with differences of expressive meaning; and they are notoriously dif­ficult to translate satisfactorily from one language to another. Even if we restrict ourselves to the medium-transferable, verbal part of utterances, we can see that a much broader range of lexical and grammatical resources than have been considered by Grice (and his followers) under the rubric of conventional implicature can be used by speakers to implicate convention­ally something over and above what they actually say.

A third, and final, point is that, just as there is no reason to limit the applicability of the notion of conventional implicature to the use of a language in more or less formal argumentation, so there is no reason to limit it to prepositional, or descriptive, meaning. I have already suggested that differences of social and expressive meaning among descriptively synonymous expres­sions (in so far as they are lexicalized in particular languages) can be brought within the scope of the notion of conventional implicature. But social and expressive meaning is conveyed at all levels of language-structure; and it is very heterogeneous. Few logicians or linguists would wish to push the notion of con­ventional implicature as far as I have done. Indeed, there are many who would deny that it has any validity at all. Some truth-conditionalists would argue that the alleged implicatures are either entailments or are implicatures of the kind that Grice called conversational, rather than conventional. Others have argued that the phenomena we have been discussing should be dealt with as cases of presupposition, but presupposition is also a somewhat controversial topic within the framework of truth-conditional semantics (and pragmatics).

For further discussion of (so-called) conventional implicature, students are referred to the more specialized works mentioned in the 'Suggestions for further reading'. In conclusion, it may be pointed out that those who have no prior theoretical commit­ment to an exclusively truth-conditional definition of'meaning' can accept that all sorts of meaning are encoded - i.e., in Grice's terms, made conventional - in the grammatical and lexical (and phonological) structure of particular languages. Of course, this does not reduce the difficulty of deciding, in the case of

 

9.5 Conversational implicatures 277

individual utterances and more generally, exactly what mean­ing is encoded in the lexemes, particles and grammatical cat­egories of particular languages.

9.5 CONVERSATIONAL IMPLICATURES

Grice's so-called conversational implicatures have aroused far more attention in linguistics than have his conventional implica­tures. I say "so-called" because the ordinary sense of 'conversa­tional' is much narrower than Grice's. We are concerned, not solely with conversations, but with all kinds of social interaction involving either spoken or written language.

The basic idea is that language-activity, most typically, is a kind of rational (and purposive) social interaction governed by the principle of co-operation. In what may now be regarded as his classic formulation of this principle, Grice recognized sev­eral kinds of co-operation which he grouped under the headings of quantity, quality, relation and manner (1975: 45-46). Each of these comprises a set of one or more subprinciples, for­mulated by Grice as prescriptive maxims, which participants normally obey, but may on occasion flout or violate.

Let us take first the subprinciple of quantity. This may be for­mulated as follows:

(24) Say as much as, and no more than, is required (in the present context and for present purposes).

Other formulations will be found in the literature, most of which (including Grice's) employ the expressions 'be [as] informative [as]' or 'make your contribution [as] informative [as]'. I have deliberately employed a slightly more general formulation, and one which explicitly mentions context. But (24) is faithful to the spirit of Grice's original and points the way to subsequent devel­opments in what may be referred to as neo-Gricean prag­matics. For the present, however, (24) may be interpreted as being equivalent to

278 Text and discourse; context and co -text

(24a) Give as much information as, and no more than, is required (in the present context and for present pur­poses).

Now, by appealing to (24) or (24a), we can account for the fact that, if x asks (of y)

Have you done the washing-up and put everything away?
and y replies

I have done the, washing up,

y may be held to have implied, in most contexts, that he or she has not put everything away. This implication, or implicature, derives from y 's presumably deliberate failure to say Yes or its equivalent to the composite proposition which is expressed in x 's utterance and is put to y for acceptance or rejection. The sim­ple proposition "I have done the washing-up" is less informative than "I have done the washing-up and I have put everything away". On the assumption that y is being duly co-operative and is being sufficiently (but not excessively) informative, x can reasonably infer that y is not able truthfully to assert "I have put everything away".

x 's assumption (in default of any evidence to the contrary) that y is being truthful depends upon x's assumption that y, being co-operative, is obeying the second subprinciple of qual-ity:

(27) Tell the truth, and do not say anything for which you have
insufficient evidence.

Once again, this formulation differs, in wording though not in spirit, from Grice's original formulation. We shall come back to the subprinciple of quality presently. Meanwhile, let us note that truthfulness - i.e., telling the truth (and nothing but the truth, though not necessarily the whole truth) - is closely allied to sincerity (which has played an important role in the theory of speech acts). It is also important to note that to tell the truth is not the same as to say what is true (i.e., to assert a set of one or more true propositions). One can say what is true whilst

 

9.5 Conversational implicatures 279

believing it to be false or not knowing that it is true. One can also say what is true with the intention to deceive, saying what one says in such a context or in such a manner, that one knows or believes that the addressee will take it to be false. It is the voli­tional, or moral, concepts of truthfulness and sincerity which underpin Grice's view of communication. These are regulated, differently in different cultures and in different social contexts, not only by sufficiency of evidence, but also by a variety of social constraints, including those imposed by the accepted, culture-dependent, norms of politeness.

Grice's subprinciple of relation has associated with it the single maxim:

(28) Be relevant.

By appealing to (28), we can impose an interpretation on the fol­lowing dialogue:

x: The clock is slow.

y: There was a power cut this morning.

In doing so, we assume that the prepositional content ofj's state­ment bears some relation to that of x's: in particular, thatj is, or might be, supplying an explanation for what x asserts to be the case. Of course, our assumption thatj's utterance is relevant to x's in this way depends not only upon our background knowl­edge about electric clocks, but also upon the further assumption that y shares this background knowledge and knows that the clock in question is, or might be, operated by electricity directly supplied from the mains. It is easy to see that such everyday con­versational exchanges as the above may depend for their co­hesion and coherence - for the property of connectedness in virtue of which we classify them readily enough as texts - upon a whole set of assumptions of this kind, specific to particular cul­tures and to particular groups.

The subprinciple of manner was explicated by Grice in terms of (at least) four maxims, as follows:

(31) Be perspicuous, by (i) avoiding obscurity of expression, (ii)
avoiding ambiguity, (iii) being brief (avoiding unnecessary prolixity), and (iv) being orderly.

 

280 Text and discourse; context and co -text

It will be scon immediately that there is likely to be a correlation, on the one hand, between being brief and giving no more infor­mation than is required, and, on the other, between being rel­evant and being orderly. The fact that there are, at least intuitively, such correlations suggests that Grice's four sub-principles can be modified and reduced in number. And this is what has happened in so-called neo-Gricean pragmatics. These later, more technical, developments are not dealt with in this book.

Much of the interest aroused by Grice's work on conversa­tional implicatures derived from its explanatory potential in respect of a variety of phenomena that are troublesome from the viewpoint of formal semantics. These include metaphorical interpretation, (so-called) indirect speech-acts, anaphoric refer­ence, and the assertion of tautologies and contradictions. Not all of these can be dealt with here. It will be helpful, however, to say something about the applicability of Grice's maxims of co-operative interaction by means of language, first of all, to the interpretation of metaphor and, then, to indirect speech-acts. This will set the context for some general comments on Grice's underlying assumptions and on the difference between implicature and other kinds of implication (and presupposition) which have been studied intensively by both linguists and logicians in the last twenty years or so. We will begin with metaphor.

We shall take as our example:

(32) 'John is a tiger',

which can be assigned both a literal and a metaphorical inter­pretation. Before considering how it might be interpreted meta­phorically in the light of Grice's principles of co-operative interaction, we must note that the point made about the literal interpretation of sentences such as (7)-(9) in section 5.2 is also relevant here.

Twentieth-century linguists of a positivist bend of mind (including many generative grammarians) have often described sentences such as (32) as being either anomalous or contradic­tory. However, provided that it does not violate conditions of

9.5 Conversational implicatures 281

categorial congruity of such generality that it could not be inter­preted, not only in the actual world, but in any possible world, sentence (32) is perfectly well-formed semantically. Moreover, the proposition that it purports to express is not necessarily false. First of all, there is nothing in the structure of English which prevents anyone from assigning the proper name 'John' to an animal, wild or domesticated (or even to an inanimate entity): this fact must always be borne in mind when English sen­tences containing proper names are under discussion, but it is not of primary concern here. More important is the fact that (32) can be given a non-contradictory literal interpretation, even if'John' is being used to refer to a person.

Indeed, there are all sorts of, socio-culturally normal, every­day situations in which 'John is a tiger' might be uttered (with reference to a man or boy) to assert a true proposition. For instance, John might be playing the role of a tiger in a play about animals. Arguably, the proposition "John is a tiger" would then be true, under a literal interpretation of 'tiger' (and also, incidentally, of the verb 'be'). Granted, there are many phi­losophers who would challenge this, but in doing so they might simply be revealing their uncritical attachment to a positivist concept of literal meaning. I mention this kind of interpretation of the sentence in question in order to show that we may not need to adjust our ontological assumptions to any significant degree in the assignment of a literal interpretation to sentences which, at first sight, might look as if they cannot sustain one. Needless to say, if we abandon the ontological assumption that the same individual cannot be (simultaneously) both a human being and a tiger (in any possible world), we can immediately assign to (32) a whole range of alternative interpretations which are non-metaphorical and, arguably, non-contradictory. As we shall see later, the fact that this is so casts doubt on the sharpness of the distinction which is drawn by many authorities between conventional and conversational implicatures, on the one hand, and between implicature and other kinds of implica­tion (including strict implication, or entailment), on the other. More generally, it casts doubt on what Quine referred to many years ago as one of the "two dogmas of modern empiricism":

282 Text and discourse; context and co -text

the assumption that there is an absolutely sharp and unchal­lengeable distinction between the analytic and the synthetic. This was mentioned above in the discussion of entailment and possible worlds (4.4). It will be taken up again, later in this sec-. tion, in connexion with the notion of the defeasibility or can-cellability of normally unchallenged contextual assumptions. For the moment, however, it suffices to note that the notion of context-independent literal meaning with which many formal semanticists operate is tacitly associated with their own con­text-dependent, philosophically challengeable, ontological assumptions.

It is also worth noting at this point that there is no closer con­nexion between literal sense and truth-conditionality than there is between metaphorical sense and truth-conditionality. If a statement is made metaphorically by uttering the sentence 'John is a tiger', the proposition thereby expressed - whatever proposition it is - will have just as determinate a truth-value as a proposition such as "John is ferocious" or "John is aggressive". Granted, there may be some indeterminacy attaching to the process of metaphorical interpretation itself: it may not be clear to an addressee which of several metaphorical interpretations he or she should assign to the utterance. But this is comparable with the problem of deciding which of the several literal senses of a polysemous expression is the one intended; and it has noth­ing to do with truth-conditionality as such.

I am not saying, of course, that all metaphorical expressions are truth-conditionally determinate, but simply that they do not differ from non-metaphorical expressions in terms of a characteristically distinctive, context-dependent, indeter­minacy. Many metaphorical statements will certainly be truth-conditionally indeterminate; and many will contain an expressive, or socio-expressive, component, which might be held to affect the determinacy of truth-value. But in this respect they are no different from non-metaphorical statements such as

(33) Mary is beautiful

or indeed

9.5 Conversational implicatures 283

John is aggressive
and

John is dynamic.

Linguists who distinguish semantics from pragmatics by means of the criterion of truth-conditionality and ascribe the meta­phorical interpretation of utterances to pragmatics tend to miss this point.

How then do Grice's maxims of co-operative interaction apply to the process of metaphorical interpretation? The general answer is, not that they guide addressees in their search for one metaphorical interpretation rather than another, but that they motivate the search itself. They enable the addressee to calcu­late (or compute) the intended meaning of the utterance as a function of its literal meaning and of the context in which it is uttered. For example, hearing or reading John is a tiger, addres­sees might reason deductively as follows, saying to themselves as it were:

(36) The speaker/writer cannot mean that literally. However, I have no grounds for believing that he/she is being unco-operative. His/her utterance has the form of a statement. Therefore, he/she must be try­ ing to tell me something, which presumably makes sense to us both (in the light of our beliefs and assumptions about the world, etc.).
He/she must also believe (if he/she is being co-operative) that I can work out the non-literal meaning for myself - presumably on the basis of the literal meaning (of the whole utterance-inscription or of one or more of its component expressions). One contextually acceptable way of using language to convey something other than what is actually said is by means of metaphor. Let me see whether lean interpret the utterance metaphorically.

I have spelt this out in some detail (though I have omitted one or two steps in the reasoning) in order to emphasize the multiplicity of assumptions that go into Gricean explanations of metaphor and other phenomena.

Let me now make explicit a few of the points that are implicit in the above account of the way in which addressees are assumed

284 Text and discourse; context and co -text

to get from the literal interpretation of an utterance to some con-textually relevant metaphorical interpretation. First, their assumption or inference that the utterance-inscription cannot have a literal interpretation does not depend upon its being semantically anomalous or contradictory: all that is required is that the literal sense should be contextually irrelevant (or improbable). Second, the whole process is subject to the con­straints imposed by the participants' beliefs and assumptions (including their beliefs and assumptions about one another's beliefs and assumptions): all communication is subject to such constraints. Third, I have included as a separate step the addres­see's recognition of the contextual appropriateness of metaphor: in certain contexts metaphor is more frequently used than in others. Indeed, there may well be occasions, determined by the socio-cultural situation or literary genre, when the use of meta­phor is so common that the addressee can skip the earlier steps in the reasoning process outlined above and start with the assumption that a given statement is more likely to be meant metaphorically than literally.

As I have said, Grice's maxims, of themselves, do not resolve for the addressee the problem of deciding upon one metaphori­cal interpretation of 'John is a tiger' rather than another. But that is not their purpose. Grice's aim was to maintain, as strictly and as consistently as possible, the distinction between what is actually said and what is conveyed (over and above, or instead of, what is said) by the fact of saying it (and not saying something else) and, at the same time, to bridge this gap, at least in prin­ciple, by showing how the application of one or more of the max­ims, by rational and co-operative addressees, to particular utterances in particular contexts of utterance, can enable them to calculate, or compute, the intended meaning. The context-dependent calculability (or computability) of conversational implicatures - their calculability being probabilistic, or heuris­tic, rather than algorithmic, or deterministic - is generally taken to be one of their defining properties. As we have seen, it is the maxim of relevance which is likely to play the major role in the metaphorical interpretation of such utterances as (32). And it will yield various results in various contexts of utterance.

9.5 Conversational implicatures 285

The way in which Grice's maxims apply to the interpretation of indirect speech-acts is, at least in principle, straightforward enough. As we saw earlier, the very notion of performing one illocutionary act indirectly "by way of performing another" is theoretically controversial (8.3). Moreover, many of the text­book examples of so-called indirect speech-acts involve the use of conventionalized, quasi-formulaic, locutions whose meaning in utterances of this kind should be regarded, from the viewpoint of synchronic descriptive linguistics, as being encoded in the language-system. It may well be that the utterance of the inter­rogative sentence

(37) 'Do you mind if I smoke?'

with the allegedly indirect illocutionary force of a request is dia-chronically explicable in terms of the notion of conversational implicature. But it is highly implausible to suggest that present-day speakers of Standard English would interpret an utterance of (37) as a request only secondarily, after having first inter­preted it as an information-seeking question. We do not need to discuss this point further in the present context.

Let us consider instead the following imaginary, but I trust realistic, bit of dialogue:

(38) x: I want to watch TV now.

y: You have notputyour toys away.

X, we shall assume, is a young child; and y is the mother (or some other person with acknowledged authority and responsibility). Each of them has uttered a declarative sentence; but neither of them, presumably, intends thereby to augment the other's knowledge of the world by making a true statement; and neither of them interprets the other's utterance as being motivated, even incidentally, by this intention. Indeed, y 's statement (and let us grant that it is a statement), if true, tells x nothing of which x is not already, perhaps resentfully and petulantly, aware. Whether x's utterance is correctly classified as a state­ment or a request (or both) is a question that need not concern us. What matters is that it is interpreted by y), in the context in

286 Text and discourse; context and co -text

which it is uttered, as a request for permission to watch televi­sion. Let us now assume that it is one of the rules of the household that x is not allowed to watch television unless and until x has put his or her toys away (or, more generally, has done a set of chores of which this is one). Given this assumption, our little bit of dialogue evidently manifests the properties of coherence and relevance. And x, being reminded of the rule and perceiving its relevance, may correctly interpretj's utterance as implicating a refusal to grant x's request.

What examples such as this demonstrate is that some, if not all, of what have been referred to in the' literature as indirect speech-acts can be plausibly accounted for in terms of the more general and more powerful notion of Gricean implicatures. This being so, there are many authorities who would seek to do away with the notion of indirect speech-acts altogether; and there are some who would question whether there is any need for illocutionary force as a distinct and identifiable part of the meaning of utterances. Recent work in pragmatics has certainly given much more prominence to implicature than it has to that of either direct or indirect illocutionary force.

These more recent developments have not been (and will not be) discussed in the present book. To conclude this section, I will explain and comment briefly upon the logical properties of Grice's conversational implicatures which are generally held to distinguish them from other kinds of implication (or presupposition). The most important of these properties is what Grice referred to as their defeasibility. To say that implicatures are defeasible is to say that their validity is context-dependent and that in particular contexts they can be cancelled without contradiction or any other kind of anomaly. For example, the conjunction of two clauses by means of and, as in

(39) John arrived late and missed the train,

would normally implicate that there is a temporal and/or causal connexion between the situations described in the two conjoined clauses. As we saw in an earlier chapter, proponents of the viewn that and-conj unction in English is truth-functional have been!

9.5 Conversational implicatures 287

able to invoke the notion of implicature in support of this view (6.3). On the assumption that anyone uttering (39) is being duly respectful of the subprinciples of manner (being orderly) and relation (being relevant), its utterance will conversationally implicate - i.e., licence the inference - that John missed the train because he arrived late (presumably at the station). The fact that this causal connexion is merely implicated, and neither expressed in what is said nor entailed (i.e., strictly implied) by what is said, is demonstrated by its defeasibility in appropriate contexts of utterance. For example, the implicature can be read­ily cancelled, without contradiction, by explicitly denying that there is a causal connexion between John's late arrival and his missing the train:

(40) John arrived late and missed the train, but it was not because he
arrived late that he missed the train [the train was delayed and did
not leave until ten minutes after he got there. So, why did he miss the
train? Maybe he did so deliberately].

There is nothing anomalous about (40) either with or (in appro­priate context) without the overtly explanatory portion of text which I have added in square brackets. And, as we saw in the preceding section, the use of but-conjunction to cancel the nor­mal conversational implicature associated with and-conjunction is explicable in Gricean terms, by appealing to the complemen­tary notion of conventional implicature. The property of defea­sibility distinguishes conversational implicature, not only from entailment, but also from conventional implicature.