|
|
|
|
Primary: Grice, H. P. (1957) ‘Meaning’, Philosophical Review, 66, pp. 377-388 (in Rosenberg/Travis (eds.) (1971) Readings in the Philosophy of Language, Englewood Cliffs, NJ: Prentice-Hall, pp. 436-444.) (1977) ‘Logic and Conversation’ (in Cole, P./J. Morgan (eds.) Syntax and Semantics, Vol 3: Speech Acts, NY: Academic Press, pp. 41-59.)
Secondary: Lycan, W. G. (2000) Philosophy of Language, London; Routledge. Devitt/Sterelny (187) Language and Reality, Oxford: Blackwell
|
|
|
Context
|
|
|
The last lecture looked at some ways in which people have tried to expand the treatment of truth conditional semantics so as to account for the effect of the context upon the interpretation of sentences. I’m sure you’ll agree it looked like a horribly unfriendly way of making such an extension, and it also seemed that even if the objections about circularity (recall: meaning is defined in terms of truth and truth is said to be defined in terms of meaning) and extensionality (recall the renates and the cordates) and so on could be satisfactorily answered, there was a good deal of language that was going to be left out of the account (questions, commands, wishes, etc.) But all these are really just objections to the truth conditional semantic approach. They claim to be no more than evidence that that approach is fundamentally unsatisfactory. They do not in themselves suggest a new approach to the problem of meaning and how language can be said to work to express meaning in the real world.
You’ll possibly also recollect that I’ve several times described the truth conditional view as being a direct successor to the Russellian theories that I introduced in the very early lectures. They too sought to interpret a sentence in terms of its supposed ‘logical form’ – a thing that was not explicitly defined anywhere by Russell, but may be thought of as being the essence of the sentence’s inferential powers expressed in some perspicuous language. In this view of things the logical form belongs to a sentence in a language. But we saw that Strawson had a cogent objection to this sort of claim. He was of the opinion that sentences themselves did not mean things, but that people could use sentences to mean things. I think this is a disagreement that marks a fundamental division in attitude towards the problems in question. I don’t quite know how to formulate the division, but it also seemed to feature in the tendency to prefer either some version of a Causal-Historical theory of naturalised meanings or some version of a meaning-as-use/Kripkensteinian interpretation of language. It’s not that the two perspectives are contradictory or necessarily incompatible, but, as I say, they do indicate two very different ways of addressing the phenomena of language.
The point of all this, of course, is to introduce just such an alternative perspective to the semantic pragmatic view. In this and the following lecture we’re going to have a look at two different specific contributions to this alternative strategy for understanding language. We’ll look at one due to Paul Grice today and we’ll talk about one inspired by Austin next time.
|
|
|
Conventional Meaning and Utterer's Meaning
|
|
|
Grice’s point of view is one that in fact arises naturally from the sorts of things that Strawson was claiming about the use of sentences earlier; and in recognition of that relationship we’re going to start by having a look at what Grice has to say about the difference between what a sentence may be said to mean just because it is the sentence that it is, and what a sentence may be said to mean in virtue of the fact that it is being used by a speaker to mean something.
I might just mention here that Grice begins his analysis by introducing us to a distinction between different kinds of meaning. In particular he makes a distinction between natural and non-natural meanings. I’ve referred to this a distinction a few times. The idea of a natural meaning is what we appeal to when we say that smoke over a hill means that there’s a fire on the hill, the alternation of light and dark rings in a tree trunk mean that the tree 1000 years old when it was cut down, and that a splitting headache in the morning means that you drank too much last night. There are various ways in which we could try to make such an idea more precise, but since it’s not really relevant to the current discussion I’m not going to say anything more about it. I just mentioned it so that you’d know where the idea came from. Just like Grice we’ll be interested only in non-natural meanings, which he divides into two kinds – conventional and utterer’s. So let’s talk about that.
Justifying the Distinction
You’ll recall how Strawson criticized Russell’s appeal to logical forms by pointing out, for example, that when someone makes a claim about the author of Waverly – as when they say that the author of Waverly is a fine stylist – Russell’s theory would insist that the statement makes several assertions; one of which is the assertion that there is one and only one thing that is an author of Waverley. Just think back to the paraphrases that we made of such statements with their declarations that ‘there is an object such that … and if any other object is like that then it is the same object and …’ and so on. If the meaning of a sentence is explicable just in terms of its logical form, says Strawson interpreting Russell, then when we make a statement that uses a sentence, the meaning of the statement is the meaning of the sentence. But Strawson thinks that this is just wrong. The statement, he says, makes no such assertion about the existence of such an object – it is no part of its role to make such a claim. Instead, we should think of the statement as presupposing that there is such an object.
In this objection we see the beginning of a more general objection that the meaning of a sentence in a statement is not to be understood purely through consideration of the features of the sentence used. It is observations of this kind that prompt Grice to suggest that there is a distinction to be drawn between the conventional meaning and the utterer’s meaning. Of course, we’re all aware – or we should be – that there’s some truth to a claim of this kind, because we’re all well aware of the many ways in which sentences can be used to mean things other than what they would be interpreted as saying in neutral circumstances. That’s still pretty vague, and the distinction will remain so, but here are some examples of cases where the distinction between conventional and utterer’s meaning are significant.
Well I’m sure you could now multiply examples to your heart’s content. You get the idea.
Utterer’s Meaning
Of course, it’d be a nicer theory if these two kinds of meaning, now that we’ve identified them as significantly different, could be shown to be unified at some deeper level. There are a number of ways that this could be done: we could show that they are both derived from an underlying notion (of meaning?) that is different from either; or we could show that one is derived from the other. In fact it’s this latter option that appeals to Grice. But this naturally begs the question: which of these is primary?
One might at first assume that conventional meaning precedes utterer’s meaning because utterer’s meaning seems to presuppose the existence of conventional meaning. Recall that when I made that weak joke about looking forward to reading your long essays the fact that I wasn’t looking forward to it was simply conveyed by the fact that the intonation with which I uttered the sentence signaled that the intended meaning was just the opposite of the conventional meaning. But if there was no conventional meaning to work on, what use would my intonations be? Similarly, in general, people use metaphors to convey meanings because the words/phrases that are used as metaphors have a conventional meaning. If the White House didn’t mean the White House conventionally it would be hard to see how it could mean the administration controlling the White House by extension. Similarly again, if ‘begging the question’ had no conventional meaning it would not be a part of the language and available for abuse, and the abuse would not always be of the same sort. But it is.
But consider further; if a million monkeys typing for a million years at some stage write out ‘the quality of mercy is not strained, it droppeth as the gentle rain from heaven’ would that sentence have some meaning? What about if they typed out ‘the quality of mercy is not strained, it droppeth as the urgle gurgle flurgle burgle’? Suppose that the wind carves the letters ‘Kilroy was here’ into the rock of the Grand Canyon. Is that a meaningful sentence? Well, consider. If it’s meaningful then someone who understands it should be able to tell us what would make it true. But there isn’t any Kilroy to whom the wind intended to refer. There is no fact of the matter of to whom Kilroy refers. So there’s no fact of the matter about whether the sentence is true or false, and I don’t know about you, but that suggests to me that there’s no meaning to the sentence – at least, not the sort of conventional meaning that we’d attribute to it if it was spoken by a speaker intending to refer to a specific someone with a specific history.
These thoughts at least suggest that the question is open. In fact, Grice comes down on the side of the primacy of utterer’s meaning, and he has a theory of how conventional meaning is to be understood in terms utterer’s meaning, but this theory is complex and we won’t look at it today. Today we shall do no more than look at what Grice thinks we can say about utterer’s meaning.
In the argument I just made to convince you that conventional meaning couldn’t exist without utterer’s meaning some appeal was made to the intentions of the language users. Grice takes these intentions to be of fundamental importance, but in a particular way. It seems that they are somewhat reflexive, we have intentions about our intentions. Grice comes to the conclusion that there is some such reflexivity by wondering how to make a distinction between ‘getting someone to think’ such and such, and ‘telling’ them the same thing. He appeals to cases such as the following.[1]
(1) I show Mr. X a photograph of Mr. Y displaying undue familiarity to Mrs. X. (2) I draw a picture of Mr. Y behaving in this manner and show it to Mr. X.
In such a case the photograph has a natural meaning that Mr. Y and Mrs. X are being naughty, and Mr. X will come to believe that that is the case no matter what he thinks about my intentions. In this case, I’m not really communicating so much as presenting evidence and letting Mr. X make up his own mind about things. On the other hand, in case (2),
it will make a difference to the effect of my picture on Mr. X whether or not he takes me to be intending to inform him (make him believe something) about Mrs. X, and not to be just doodling or trying to produce a work of art.
Basically, therefore, we think that the utterer’s meaning depends upon the intention of the utterer to communicate in virtue of the audience’s recognition of the utterer having that intention. But this is a very imprecise formulation. More precisely Grice makes an initial proposal that
‘A meantNN something by x’ is (roughly) equivalent to ‘A intended the utterance of x to produce some effect in an audience by means of the recognition of this intention’; and we may add that to ask what A meant is to ask for a specification of the intended effect.
But for various reasons this isn’t quite what he wants either; for example, not just any effect will do. At this point the Gricean research program becomes obsessed with further and further refinement of the basic Gricean formulation of utterer’s meaning intended to account for more and more convoluted situations. Just for an example, here’s a version from 1969 – not the most evolved version.
‘U meant something by x’ is true iff U uttered x intending thereby: (1) that A should produce response r, (2) that A should, at least partly on the basis of x, think that U intended (1), (3) that A should think that U intended (2), (4) that A’s production of r should be based (at least in part) on A’s thought that U intended (1), (5) that A should think U intended (4).
Remarks on Utterer’s Meaning
1. You should notice that the effect that is supposedly produced in A is not necessarily the adoption or reinforcement of a belief, because we can tell things to people not intending in any way to affect their beliefs. For example, after the end of the last tutorial, a few students were talking to each other heatedly about the recent election results – but my impression was that they all believed all the same things, and all knew that the others believed all the same things. And yet it looked as if they were quite satisfied by the conversation they were having. So what was the point of the conversation at all? Not belief fixing at any rate. There are similar problems with people answering test questions.
2. More than this, it’s clear to anyone that the things that we say to ourselves when nobody is around have meanings. If I’m driving around in my car, I’m usually muttering to myself things like ‘Get out of 2nd gear you dribbling loon.’ or ‘How about an indicator, dipstick?’, or other things of that kind. There’s no audience that I’m trying to influence or create an effect in, so according to Grice’s rules these things are meaningless. I reject that conclusion. And Grice accepts that this isn’t right too. He thinks that this indicates that the definition needs to be refined to talk of hypothetical audiences. For example, the last statement I gave for Grice’s position would be revised to say that
‘U meant something by x’ is true iff U uttered x intending thereby that if there were an audience A: etc…
3. The complexity of Grice’s definitions of what it is to non-naturally mean something is somewhat perplexing. It seems to be a result of a particular conception that he has of what a philosophy of language should aim to accomplish. In his work he is trying to provide an analysis of a commonplace concept so that it can be entirely explained in terms of other concepts. It is as if he were trying to treat ‘meaning’ in the same way that we might treat ‘bachelor’. The definition that we could give for ‘bachelor’ would be necessarily true and can be known a priori, and this is what he aspires to for a definition of ‘meaning’. The definition has to be such that it exactly captures the intuitive judgments that we make of ‘meaning’ in every possible situation. Of course, this means that he can’t admit that the notion of ‘meaning’ might be some incoherent notion that is a part of our ‘folk theory’ of language but not part of a theory that accurately represents reality. Not everyone agrees that this is the best way to approach philosophical questions of this sort.
4. Furthermore, are we supposed to believe that some complex formulation like that has some sort of psychological reality? Does anybody here have anything at all like that consciously in mind when they set out to communicate to someone else? Inconceivable! Luckily, Grice does not make any such claim. Indeed, he specifically rejects it. He appears to believe that these intentions and beliefs and other attitudes can be considered merely behaviouristically, but we don’t have to follow him in that claim. Instead we could simply consider that, whatever these attitudes are they are merely subconscious or tacit.
5. There’s a problem example I’d like to mention without going into any analysis of it, just because it’s a fairly well known one. The example imagines that there’s an American soldier fighting in Italy in WWII. He’s captured by Italians and wants to trick them into believing that he’s really a German officer. They don’t know German and he only knows one line from a poem he learnt at school. So he says to them
Kennst du das land wo die Zitronen bluhen?
Which they interpret as him saying ‘Hi, I’m a German officer’ but it hardly seems likely that when the American said what he said he meant that he was a German officer. Seems like something else is going on there.
[1] Grice (1957) p. 440.
|
|
|
Implications in Natural Language
|
|
|
You’ll recall that we started this discussion of the distinction between utterer’s meaning and conventional meaning by referring back to Strawson’s critique of Russell, and in particular to the claims that Strawson made about what sorts of entailments are allowed to be read into the things that we say. We return to this theme now, for Grice has much to say about how some sentences are implied by other sentences in certain contexts. Strawson had made the point that many entailments which were a necessary consequence of the logical form of a sentence were actually more than our intuitions would allow for natural language. He specifically said, you’ll remember, that to claim that ‘the present king of France was bald’ did not entail that there was a present king of France- contrary to the Russellian claim. Now Grice makes a complementary claim that there are relations of entailment (using the word non-technically) that exist in natural language, under the usual conventions, that are not captured by the logical form of the sentences. If this is correct, then there is even more reason to believe that the commonsense notion of implication is only an approximation to the logical notion. Of course there are several ways to react to that: one could say, well we already knew that material implication wasn’t like real implication, it’s only the most important part; or one could say, logical implication is the only real implication and all other things that we call by that name are parasitic upon that notion and are to some degree incoherent folk notions that must be discarded; or one could say that we just need to work harder to see how these other types of implication can be reduced to real implication; or …. My point is just that the story doesn’t end even if you think that Grice is correct in what he says about informal implications, which shouldn’t be a surprise to anybody at this stage.
In any case it’s pretty clear that we’re going to need some way of labelling all these other-than-logical types of implication. A word that Grice uses is implicature. Let it be so understood from now on.
Semantic Presupposition
One of the forms of implicature that Grice talks about is of course this presupposition that Strawson mentioned. In the example that Strawson used, he claimed that because there was no King of France, any statement about the King of France could be neither true nor false. The utterer would have simply failed to make a statement. We could try to make that sort of thing a marker for this type of implicature. Thus:
If S presupposes P then: If S is true then P is true, and If P is false then S is neither true nor false (and neither is not-S.)
Another example of this – if you’re tired of the king of France, and who wouldn’t be – is:
S1 [Student name]’s second essay is good.
which seems to presuppose that:
P1 [Student name] has submitted a second essay.
You can quite easily see that if it was true that [student name]’s second essay is good, then it would naturally follow that [student name] has submitted a second essay. On the other hand if [student name] has not submitted a second essay – and they haven’t, because it’s not due yet – then it may be neither true nor false that [student name]’s second essay is good. And it may be neither true nor false that [student name]’s second essay is not good.
And this sort of thing doesn’t just apply to indicatives. There are also presuppositions to be found in interrogatives, imperatives, optatives, and so on. Just consider the famous old taunt:
Q2 Have you stopped beating your wife?
This seems to presuppose that:
P2 You are beating your wife.
The marker for this sort of thing can’t be just as we described for the indicatives but there’s an easy modification that we could make that would work, I think:
If the question Q presupposes P then: If one can properly answer yes to Q then P is true, and If P is false then one cannot properly answer Q either yes or no.
We all know why you can’t answer yes if P2 is not true: it’s because you can’t stop doing something unless you have been doing it. It’s a little trickier to say why you can’t properly answer no, but it seems to be that the presupposition is understood by the audience by default and unless you specifically deny the presupposition the audience will understand that, yes, you do beat your wife, and no, you have not yet stopped doing so, which is not a good thing.
I leave you to make the appropriate modifications for the imperative:
O3 Stop beating your wife!
and the optative:
W4 If only you would stop beating your wife.
Conveyed Meaning
Another way that some sentences seem to be implicated by other sentences is more closely related to Grice’s idea of utterer’s meaning. There are certain ways of stating things that invite the audience to understand much more than the conventional meaning of the sentence uttered would justify. An example of such a thing is what you occasionally get in school reports:
For Maths: Johnny likes using the coloured chalk.
From which we are entitled to infer that Johnny can’t add.
Or there are job references:
For System Administrator: John always dressed very nicely.
From which we may infer that John is not a competent worker.
In such cases it is clear that the failure on the part of the utterer to say the good things that are expected in this situation are an invitation to conclude that the things may not be uttered by someone wishing to speak honestly. Thus the audience will conclude that the things they would expect to hear of a good student or a good candidate would be false if said of Johnny or John. This is a particular case, probably of conversational implicature, which we’ll look at in just a moment. It is the implicature most closely associated with Grice’s name.
Invited Inference
Yet another form of implicature can be seen when statements which do not explicitly state an implicative relationship are such that such a relationship is naturally assumed to exist. Very often we join two statements together and the mere fact that we think that this conjunction is worth making will suggest to our audience that there is some sort of relationship between the two statements; and as often as not the relation that will be supposed to exist is one of cause and effect. Compare for example:
The door slammed and the cake fell.
It’s natural to think that the cake fell because the door slammed, although the sentence says nothing about this. But this is an implicature that is assisted by our remembrance of the sorts of things that have made our cakes fall in the past, so perhaps it’s not an entirely fair example. Very well, how about:
The baby smiled and Frank shuddered in horror.
We infer that Frank shuddered in horror because the baby smiled. And yet we don’t have much in the way of experiences that would suggest that the smiling of babies is naturally causally connected with feelings of revulsion and horror. This inference seems to go against all our intuitions about how babies smiling affect the world, and indeed the sentence itself only states that the two events occurred, without making any claim about how they are related. The only reason we seek some relation between them is because we assume that the utterer must have meant something by putting the two statements together in that way. But this sort of sensitivity to general conversational rules brings us to Grice’s most developed version of implicature, which he thought bid fair to explain almost all the implicatures that he’d recognised in natural language use.
Conversational Implicature
Grice thought that much of what passed in conversation could be understood in terms of social conventions by which the normal function of conversation could be facilitated; and the normal function of conversation was quite simply the communication of certain ideas from an utterer to an audience. He actually identified a range of conventions, which it is worth our while noting. They are his conversational maxims, and they include such insights as:
M1. The Maxim of Strength: Make your contribution to a conversation as informative as required. (For example, if someone pulls up to you in Toowong and asks what direction is West End, you shouldn’t simply point towards West End, because they are clearly asking what direction they should travel to drive to West End, and if they follow your finger they’ll wind up in the river.)
M2. Do not make your contribution more informative than is required. (For example, if someone asks you how to get to West End, it is sufficient to give them instructions for just one route. There is no value in giving them a range of possible routes and delaying them for half an hour while you enumerate all the possibilities.)
M3. Do not say what you believe to be false. (Note that saying what you believe to be false is different from just saying what is false. This seems to be a difference that people have difficulty recognising. The maxim is saying that one shouldn’t lie, not that one is obliged to be omniscient. The former is a commandment of every moral system I know of, whereas the latter is something that only God can achieve.)
M4. Do not say that for which you lack adequate evidence. (You might think that thins is a little bit of an excessive demand; after all, who’s to say what is to count as ‘adequate’ evidence. Clearly it can’t mean that we can never say anything that we’re not able to prove beyond all reasonable doubt. How many of us here would be able to say, under that condition, that the moon went around the Earth, or that people evolved from non-people, or that matter is made of atoms. No, what this condition is supposed to prevent is the sort of thing that you get in arguments where someone just makes up facts – usually in the form of amazing statistics – which they think is probably right, and which supports their side of the argument. Did you know that studies have shown that 78% of all statistics cited are completely bogus?)
M5. The Maxim of Relevance: Be relevant. (For example, if someone asks you the direction to West End, don’t start giving them a lecture on diseases of the blood. It’s not relevant to the conversation and will disturb the audience, who will look for reasons for why what you are saying is relevant.)
M6. Avoid ambiguity. (There’s a classic example of ambiguity in a story from Herodotus.
The Lydians … were instructed by Croesus to ask the oracles if he should undertake the campaign against Persia. … To this question both oracles returned a similar answer; they foretold that if he attacked Persia he would destroy a great empire.
Well he did attack Persia, and he did destroy a great empire. Unfortunately it was his own empire. In this case the oracles were not cooperating in the conversation.
M7. Maxim of Brevity: Be as brief as possible. (There’s no reason to spend more time making a point than is absolutely necessary. The mere fact that you are taking longer to say something than is necessary will itself be taken as something significant by the audience.)
These maxims are all supposed to be derivable from a more general principle of communicative action which Grice calls the Cooperative Principle:
CP: Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk-exchange in which you are engaged.
Which might be further compressed to just ‘say the appropriate thing at the appropriate time.’ The fact that Grice is able to use his cooperative principle to develop those more informative maxims is evidence that there is something useful being said in the principle, and that it’s not completely trivial. We can use this cooperative principle to explain how we derive information from the communications of our conversational partners, but only if we’re prepared to accept that something like the following reasoning occurs when we understand q from the conversation p:
CR: He has said that p; there is no reason to suppose that he is not observing the maxims … ; he could not be doing this unless he thought that q; he knows (and knows that I know that he knows) that I can see that the supposition that he thinks that q is required; he has done nothing to stop me thinking that q; therefore he intends me to think, or is at least willing to allow me to think, that q; and so he has implicated that q.
Some have thought that it’s entirely unlikely that we are able to make all these deductions in real time – especially when one considers that while this is going on we’ve also got to be dealing with his equally complex process of determining the utterer’s meaning of the sentences that occur in the conversation. I have to say that I don’t see very much problem with this, but perhaps you do.
It’s also worth remarking upon one condition in the reasoning sketched above, that the speaker has done nothing to stop me thinking that q. The significance of this is that Grice thinks that all implicatures are such that they can be blocked. Some added piece of information in the conversation can remove the possibility of seeing q as being implicated by the speaker. This is quite different from normal implication, because one you have, for example, that ‘A or B’ and that ‘not A’ then nothing you can add will prevent the implication that ‘B’.
Now, this principle is supposed to be a norm that conversations are shaped by. The normative conversation is one in which the principle is observed scrupulously. But, being the clever conversationalists that we are, we also find a good deal of communicative value in deliberately violating the norm. In particular we have found value in violating the norm in those situations where we have given our conversational partners sufficient reason in the course of the conversation to believe that we are aware of the obligations of conversation, so that they will notice and remark upon our violation of the maxims. Some of those uses I’m sure will have occurred to you while I was listing the selection of maxims. Look how we can use the violation of those maxims to our advantage.
~M1. Violating the maxim of strength: The previous example I gave of the uninformative job reference is one way in which the failure to give as much information as might rightly be expected can itself be informative. It is because John’s resume fails to tell us things that we can conclude other things; i.e. that he’s no good.
~M2. Making your contribution more informative than is required: Doctors get this a lot. You’re at a party and a doctor asks ‘how are you?’ If you answer at great length about your aching bones, and the funny twinges that you get just after your squash games, then the excess of information will be taken by the doctor as a request for a free consultation. But you can’t ask straight out for that because that’d be rude.
~M5. Violating the maxim of relevance: When someone – perhaps a casual acquaintance – asks you about some renovations that you’ll be undertaking, you may decide to give them full details of the time and place and required equipment for your working bee next week end. They will wonder why you have told them these things. They will wonder what relevance this information has to them or the course of the conversation. Perhaps they will think, ‘Oh my God. Surely he doesn’t expect me to volunteer to come over this week end and help him put up his outdoor spa? I think it’s time I was getting home.’
Conventional Implicature
There is a final form of implicature that Grice proposes and that I think is worth mentioning here, and that is something called conventional implicature, which is distinguished from conversational implicature only in that the implicature does not have to be reasoned out in the fashion of (CR) but is known immediately. This distinction is interesting because it indicates that Grice did not have in mind a behaviouristic interpretation of the template of conversational reasoning: it was supposed to be a real description of what we knowingly do in our heads while we’re having a conversation.
An example of conventional implicature that Grice (almost) uses is:
He is an Englishman, but he is not brave
The implicature is clearly that Englishmen are all brave. |
|