Chapter 7
Towards a Theory of Meaning
7.1 Introduction
Having considered and found fault with a number of attempts to provide an adequate theory of the actual-language relation, I will now make a number of proposals concerning some of the background assumptions of the pursuit of such a theory. One can suppose that previous theories have failed because there just can't be an adequate theory of the actual-language relation, that is, one can suppose that somehow something like what Wittgenstein thought was right and that theories of meaning are impossible. But I really don't know of any good reason why this has to be so. So I prefer to believe that there are solvable problems with some of the assumptions that have been in the background of past attempts at theorizing about meaning. What I will do in this chapter is say what I think some of these problems have been and how I think that these problems might best be solved.
In section 7.2 I will discuss what I think a Gricean needs to do to escape the sort of counterexample that Loar found for his own theory of meaning and that I discussed in section 5.5 above as well as in section 6.7 where I showed that this counterexample is also a serious problem for Schiffer's theory. In essence I will propose a notion of saying to replace the more common notions of speaker-meaning that Gricean's have traditionally favored. My notion of saying will require that a certain common conception be dropped of some of the requirements of Grice's theory of implicature. But I will argue that, on the change of conception required by adopting my proposal concerning saying, nothing in Grice's theory of implicature is lost and that if there is a gain for the theory of literalexpression-meaning, then we do well to adopt my proposal along with the requisite changes in the conception of the theory of implicature. In section 7.3 I will talk about a certain presupposition concerning the magnitude of natural languages that I think is responsible in part for forcing theorists to accept the disjunction discussed in chapter 5 of founding infinitudes of meaning-facts either in finite sets of facts about actual usage or in CMTs. I will argue that there is no reason to accept this presupposition and that it should therefore be dropped. In section 7.4 I will present what will appear a further serious counterexample to Schiffer's theory discussed in chapter 6. But I will argue that this counterexample is only a counterexample given a false, if widely held, presupposition about the aims of the theory of meaning. I will then argue that this presupposition is yet another that should be dropped. Finally, in section 7.5, I will say what I think a theory of meaning should look like given the revisions I have suggested to the presuppositions that have been a standard part of theorizing over the past 35 or so years.
So the main goals of this chapter are to loosen up the belief in a number of widely held presuppositions that have led to problems for the theory of meaning in the past and also to suggest what an adequate theory of meaning might look like.
7.2 Saying
In section 5.5 I discussed a counterexample Loar presented to a version of his own theory. Loar noticed that on his proposal for a theory of the actual-language relation, given some plausible assumptions about English speakers, the English sentence "Apollo set the evening sky ablaze" means among English speakers something like that the sunsetis intensely beautiful this evening rather than that Apollo set the evening sky ablaze. Loar diagnoses the problem with his theory to be its lack of an adequate consideration of the compositional structure of sentences. But I believe the real problem with Loar's theory is that the notion of speaker-meaning he employs is too strong a notion to be useful in a theory of expression-meaning. I will now propose a slightly weaker notion to replace the Gricean notion of speaker-meaning in attempts to provide a theory of expression-meaning. My notion will be one of saying, and though there will be some relation between my notion of saying and the notion expressed by the English word "saying", the point now is to engineer a notion that is suitable to the theory of expression-meaning, not to analyze an ordinary one.
The notions of speaker-meaning that Grice and the theorists in the Gricean tradition have employed all have it, more-or-less, that to speaker-mean is to intend to get a listener to believe the proposition that is speaker-meant. For convenience, I will talk from now on as if there is just one proposed analysis of speaker-meaning floating around and that this is one of its essential features. I want to oppose to this notion a notion of saying, where what is said need not be what the speaker intends to get an audience to believe, even if saying always is an attempt to get an audience to believe something. To say something, that is, is to try to get an audience to believe something by trying to get them to consider what is said. Sometimes a person speaker-means what is said, but sometimes not. Thus, the notion of saying I have in mind can be seen as comporting nicely with the ordinary way of speaking whereby a person can say what they mean and mean what they say, but need not do so. So, I can say to you that Apollo set the eveningsky ablaze and not mean it, that is, not mean that you come to believe it, or even believe that I believe it. I say it to have you consider it and recognize, partly on the basis of your considering it that I want you to believe something else, namely, that there is an intensely beautiful sunset this evening.
My proposal, then, is that the following notion of saying is more suitable to a theory of expression-meaning than the more common notion of speaker-meaning.
[S]A speaker S says a proposition by uttering a sentence just in case there is an audience A such that
(I)for some proposition '', S intends that
(1)A comes to believe '',
(2)A comes to consider whether to believe ,
(3)A recognizes S's intentions (1) and (2), and
(4)A comes to believe '' by reasoning at least partly from A's recognition of S's intentions (1) and (2),
and(II)there is no proposition ', different from , such that S intends that
(5)A comes to consider whether to believe ',
(6)A recognizes S's intention (5), and
(7)A comes to believe that S intends A to consider whether to believe by reasoning at least partly from A's recognition of S's intention (5).<1>
A central insight of Grice's is that a speaker's intention to get an audience to believe a proposition can itself count as evidence for the audience that the proposition is true. I preserve this insight in my analysis of saying. To say is to intend to get an audience to believe something and to get an audience to believe that thing on the basis of their recognition of the speaker's intention to get them to believe it. But, the thing that a speaker wants an audience to believe does not have to be the thing the speaker says. The speaker will intend the audience to consider a certain proposition and, on the basis of this consideration and recognition of the speaker's intention to bring it about, to infer what proposition the speaker actually wants the audience to believe. Thus, I might utter "Apollo set the evening sky ablaze" intending that you understand that I am trying to get you to believe something and trying to get you to believe that particular something on the basis, in part, of your considering the truth of the proposition that Apollo set the evening sky ablaze. I don't want you to believe that Apollo set the evening sky ablaze, or to believe that I believe it. I just want you to consider it and reason from your recognition of my wanting you to consider it to a recognition of what proposition I actually do want you to believe, namely, that the sunset is intensely beautiful this evening. Then, once you have grasped that I really want you to believe that the sunset is intensely beautiful this evening, I want you to reason, in the usual Gricean way, from your recognition of my intention to get you to believe that the sunset is intensely beautiful this evening to the belief that the sunset is intensely beautiful this evening. I say that Apollo set the evening sky ablaze as a part of my speaker-meaning that the sunset is intensely beautiful this evening.
Thus, I intend to get you to recognize my intention to get you to believe that the sunset is intensely beautiful this evening and thereby believe that the sunset is intensely beautiful this evening by getting you to recognize first that I want you to consider whether to believe that Apollo set the evening sky ablaze. I know, let's say, that you will immediately upon consideration reject it's being the case that Apollo set the evening sky ablaze because I know that you know that Apollo is a mythical god. I also know that youwill immediately reject the thought that I believe that Apollo set the evening sky ablaze because I know that you know that I know that there is no such thing as Apollo. But I expect that you will try to make sense of my having tried to get you to consider this belief about Apollo in the context of your recognizing that I do want you to believe something. And I expect you to figure out, given all of this along with, let's suppose, some general conversational principles and other background assumptions, that what I am trying to get you to believe is that the sunset is intensely beautiful this evening.
In the case where I mean what I say, the mechanism is in general the same, but I intend that you figure out that the proposition that I want you to consider is also the proposition that I want you to believe. This explicitly adds a step to the story that Grice suggested, but ultimately the mechanism is the same: an audience figures out on the basis of contextual information that my interest in having them recognize my intention to have them consider a proposition is, in fact, to have them come to recognize my intention to have them believe that proposition and thereby to come to believe it. And I believe that it is not implausible that this is what does happen even in the simplest cases of saying.
My story doesn't contradict Grice's story about speaker-meaning, but makes explicit a step that I believe was implicitly there in the first place. For Grice never says what mechanism or reasoning process is used or intended by a speaker when attempting to get an audience to recognize the speaker's intentions. I have found it helpful to make part of such intended mechanisms or reasoning processes explicit. Namely, in cases of speaker-meaning a speaker intends to have an audience come to recognize the speaker's intention to get the audience to believe a proposition by reasoning in part from theirrecognition of the speaker's intention to get the audience to consider whether to believe that proposition.
Notice that if I intend (a) that you consider a proposition ', (b) that you recognize this intention, and (c) that you reason from this intention to the belief that I want you to consider the proposition , then I haven't said to you, even if I might have suggested it. It seems that in that case I have said ' to you intending that you consider another proposition, , and thereby come to adduce what I want you to believe. Consider the following case cited by Grice:
Take the complex example of the British General who captured the town of Sind and sent back the message Peccavi. The ambiguity involved ('I have Sind'/'I have sinned') is phonemic, not morphemic; and the expression actually used is unambiguous....<2>
In this case it seems that the General has said that he sinned, and he said this because he knew that his listener would come to consider the proposition that the General had captured Sind and go on from there ultimately to the belief that the general had captured Sind. It seems equally clear that the General did not say that he had captured Sind. Clause (II) of [S] is meant to help capture this latter fact. The speaker invites the listener to consider one proposition in order to go on from there to consider another, and then, ultimately, to go on to figure out what the speaker intends to get the listener to believe. But the second proposition considered here is not said by the speaker and an analysis of saying should predict this as mine does.<3>
Thus, the motivations for the various parts of [S] should be clear. It should also be clear how a theory of the actual-language relation that used such a notion as saying can avoid the Loar-style problem discussed above in section 5.5. For, if a theory of theactual-language relation does not work by correlating utterances with what their utterer's intend to get their audiences to believe but by correlating utterances with what their utterer's primarily intend to get their audiences to consider, then the possibility that a sentence might never be uttered by a speaker to mean what it means need not matter. In the Loar case, the (supposed) fact that English speakers would tend to speaker-mean by "Apollo set the evening sky ablaze" that the sunset is intensely beautiful this evening forced his theory to identify that proposition as the meaning of the sentence. But English speakers would say that Apollo set the evening sky ablaze by uttering that sentence whatever they might mean. And it seems intuitively right to say that sentences mean what they can be used to say, not what speaker's intend to get their hearers to believe.
In Grice's "Logic and Conversation" Grice says the following about the notion of saying that he employs in discussing his theory of implicature:
In the sense in which I am using the word say, I intend what someone has said to be closely related to the conventional meaning of the words (the sentence) he has uttered.<4>
I am not entirely sure what Grice took the close relation to be between what is said and what is conventionally meant. But I believe that the relation between the notion of saying required by Grice's theory of implicature and the notion of conventional meaning is not a conceptual relation, that is, that the notion of saying required by Grice's theory of implicature does not metaphysically require conventional meaning. My notion of saying, I submit, can be used in the theory of implicature without any loss of explanatory power for that theory. My notion of saying does not require that utterances used for saying conventionally mean what they are used to say. It does turn out, in actualsituations, that we use the conventionally meaningful as devices for saying, but this is not of metaphysical necessity.
It is certainly conceivable that one can say a proposition by uttering a sound that has no conventional meaning and that one can do this in order to implicate some other proposition. Suppose that "Grrr" has no conventional meaning between you and I. But suppose that somehow I have reason to believe that if I utter the sound "Grrr" you will infer that I am trying to get you to consider whether to believe that I am angry. And suppose that I know that you know that I am not, in fact, angry. I want you to believe that a certain friend of ours is angry. So I utter "Grrr" intending to get you to consider whether to believe that I am angry, knowing that you know that I am not, and intending you to reason from your recognition that I intended you to consider whether to believe that I am angry along with certain facts I know to be known to you about our friend, that I am actually trying to get you to believe that our friend is angry. And I intend for you to reason from your recognition of my intention to get you to believe that our friend is angry to the belief that our friend is angry. Thus, in this case I haven't said that our friend is angry. Rather, I said that I was angry and I implicated that our friend was angry. And I did this without using any conventionally-meaning expressions.
It may be that for us, such a story as this is unlikely to occur and sounds a bit awkward for that reason. But, such a story is conceivable, and that is all that matters here. Neither saying nor implicating requires the use of conventional devices.
Now, it does turn out that conventionally-meaning expressions will generally be more suitable for generating implicatures because they are more dependable as devicesfor getting people to consider whether to believe propositions. And this dependability is a way that saying ends up, as Grice suggested, closely related to using conventionally-meaningful expressions. But the connection is no closer than this: when we want to get someone to believe one thing by saying another perhaps we, as a matter of fact, always use conventional devices; but we don't have to. Grice's saying that saying is closely related to conventional meaning might be suggestive of a true identification given only our actual practices, but it is not conceptually necessary. There is no reason that the psychological mechanisms involved in creating implicatures should not work for communicators who have no conventional devices at their disposal.
In fact, I believe that Grice's theory of implicature even requires a more substantial notion of saying than merely uttering an expression that conventionally means. For when I implicate something, I do so, according to Grice's view, by intending an audience to consider the proposition that I say. I know that my audience will reason a certain way about that proposition, and I, in fact, intend them to do so. It is through recognition of such a complex intention concerning their reasoning and the proposition that I suggest they consider that I expect that I can succeed in getting them to understand what I am trying to implicate. But this complex intention is not merely the intention to utter a sentence that conventionally means something. Rather, I will sometimes utter a sentence that conventionally means something because I have this complex intention. And I believe that my notion of saying characterizes the complex intention required here.
It is hard, then, to see how a theory that tries to identify expression-meaning as some sort of correlation between sentences and propositional-speech-acts can avoid theLoar-style problem if the propositional-speech-act it centrally employs is the traditional one of speaker-meaning, that is, of a speaker intending to get an audience to believe what is said. Thus I believe that my notion of saying is the more suitable notion and should be adopted for the theory of expression-meaning. This affects the theory of implicature somewhat. But the idea that the theory of implicature requires a notion of saying that is less robust than the notion that I have glossed is mistaken, as I have argued. Adopting my notion of saying may have a slight affect on the general way that one looks at the theory of implicature, but not on anything of substance in it. So there are no reasons not to adopt my notion of saying that stem from considerations of the theory of implicature.
7.3 The Unusable
If a theorist believes that, in fact, the number of meaningful sentences for communities of natural-language users is infinite, then that theorist will be faced with the task of saying how this infinitude of meaning facts can be possible for finite creatures such as us. Most attempts to do this have involved either supposing that some method can be found for basing the infinitude of meaning facts on facts about the finite corpus of actually uttered sentences or on relations in which language-users stand towards CMTs; some theorists have even tried both strategies at once.<5> The former strategy seems to entangle the theorist in truthfulness-by-silence-style problems, while the latter obligates the theorist to the existence of CMTs. I believe that neither of these options is viable for reasons that I have already discussed. And I know of no other way of understanding infinitudes of meaning facts for finite beings. So, I believe it is time to re-examine thepresupposition that the number of meaningful sentences for communities of natural-language users is infinite.
An important background idea that helps one to believe that natural languages are infinite in magnitude is Chomsky's famous competence/performance distinction. It is supposed that though we cannot process really large sentences, this is merely a fact about our actual abilities in performance, not our competence; our linguistic competence, so it is believed, is not limited by such facts about our actual performance. There is supposed to be an abstract level of description whereby, though we cannot use certain sentences, still, these sentences remain sentences of our language. And if there are to be infinitely many sentences in natural languages, then most of the sentences of natural languages will be unusable. So, I claim, behind the thesis that natural languages in fact contain infinitely many meaningful sentences lies this view that sense can be made of a notion of competence with unusable sentences. In the following I will argue that there really isn't a very good sense that anyone has ever made of this notion of competence with unusable sentences. For the idea that there can be such competence simply is the idea that there can be unusable but meaningful sentences. And note that my attack on the notion of competence-with-the-unusable is not an attack on the notion of competence-with-the-unused: there will surely be many many - if not infinitely many - sentences that we do have competence with, but which we simply have never and perhaps will never use. But whereas I believe that no good sense can be made of the notion of competence-with-the-unusable, I do believe that there is perfectly good sense to be made of the notion of competence-with-the-unused.Many, if not most or even all philosophers, who have concerned themselves with such matters have taken the view quite seriously that there can be unusable but meaningful sentences. And this supposition has constrained theorizing about meaning as I indicated above. In fact, the only philosopher I know who has ever even suggested the possibility of a theory of meaning that denied that unusable sentences could be meaningful is Brian Loar. But as quickly as Loar suggests this possibility, he attempts avoiding it:
It is not wholly far-fetched to claim that the incomprehensibly complex English 'sentences' are not really part of our language at all - after all, we cannot understand them! If English were thus reduced to a finite fragment of what it would be if our brains were larger, there would be no problem for [a certain reading of a theory Loar suggests]. But that is a wildly controversial solution - and an avoidable one.<6>
I believe that the claim that the unusable lacks meaning is correct, even if it appears, as Loar says, "wildly controversial". But some have thought that at least some unusable sentences are meaningful. I believe that for the past 40 years or so the main source of this view has been a certain argument that seems to be suggested by Chomsky in various places - I don't know whether he ever has made an explicit such argument. Here is a passage that suggests this argument I have in mind from Chomsky's earliest publication on transformational grammar:
We might arbitrarily decree that such processes of sentence formation in English as those we are discussing cannot be carried out more than n times, for some fixed n. This would of course, make English a finite state language, as, for example, would a limitation of English sentences to length of less than a million words. Such arbitrary limitations serve no useful purpose, however.<7>
What exactly Chomsky intends to be arguing for in this passage is a debatablematter. But, the argument that I think many philosophers have read into such considerations of Chomsky's goes something like the this: "if you only consider the data that comes from the 'intuitions' of natural-language users concerning the languages they speak, it will be impossible to draw a principled line that marks off where sentences become too complex to be processable; there is, therefore, no empirically motivated line to be drawn in characterizing linguistic competence and, therefore, we shouldn't draw one; so even really large sentences are to be considered meaningful if they are constructed from the syntactic rules that seem best to explain the finite set of intuitions actually considered in constructing a syntactic theory."
If this is an argument that Chomsky endorses, it seems to create a tension with views he later proposes concerning the arbitrariness of restricting the data which the grammarian in principle may use in constructing a theory to the 'intuitions' of natural-language users:
In principle, evidence ... [for the linguist] could come from many different sources apart from judgments concerning the form and meaning of expressions: perceptual experiments, the study of acquisition and deficit or of partially invented languages such as creoles, or of literary usage or language change, neurology, biochemistry, and so on. ... As in the case of any inquiry into some aspect of the physical world, there is no way of delimiting the kinds of evidence that might, in principle, prove relevant.<8>
If you believe that what is being investigated by the grammarian is a part of nature in the way that Chomsky seems to be suggesting here, it is hard to see how a conclusion about the potential complexity of sentences that is drawn from mere consideration of the corpus of 'judgments' or 'intuitions' of natural-language users can be held in high regard. If it seems doubtful that consideration of the corpus of everyday intuitions that we have aboutthe physical world would lead to a true physical theory, it should similarly seem doubtful that consideration of the corpus of intuitions of natural-language users about their language will lead to a true theory of that language. Perhaps sometimes extrapolations from such a corpus will lead to a truth about language, but it seems that if you believe language to be a phenomenon akin in relevant respects to physics or some other natural science, then you will have some healthy doubt about conclusions that you draw on the basis of people's intuitions. It may be that there are no limits on sentence-complexity discernable given only user intuitions. But given further data it could well become apparent that there are very definite limits on the grammar employed by a natural-language user. This is not something that can be determined ahead of empirical investigation.
Pressing the analogy with physics is interesting here. For it may have seemed not too long ago - perhaps even now still - that the physical world according to the "simplest" summary available of people's actual physical intuitions is such that it is possible to continuously extend a line without it's ever intersecting with itself. But, aside from questions about the notion of a straight line, it is understood that whether this is possible depends on complex issues about the density of matter in the universe or some such matters. It might be countered here that in the absence of further data, extrapolation from the corpus of intuitions would be the only rational means of guessing what the universe is actually like. So in previous days, conclusions drawn from the corpus of everyday physical intuitions were the rational conclusions to draw, however problematic they might be: it seems I am merely pointing to the underdetermination of theory by a finite set ofdata.
The fact is that in the language case there already is additional data available. We know that we are finite creatures and that our processing abilities are finite. To extrapolate from the finite corpus of intuitions about particular sentences to conclusions about the potential complexity of any sentence whatsoever will have to be informed by this finiteness somehow.
Chomsky tries to accommodate infinite complexity in languages that are used by finite creatures by invoking the competence/performance distinction. But he doesn't tell us what competence amounts to: to do this is to give a theory of meaning in some sense close to that which is the topic of this dissertation. Chomsky doesn't do this. There is no theory of competence without a theory of meaning or ahead of a theory of meaning. To simply say that there is is to beg really important questions.
Seen in this light, the argument I gloss above clearly seems fallacious. This was originally pointed out by Ziff:
But where is the line between the deviant and the nondeviant to be drawn? Why think one has to draw lines? And to see that after all one needn't...one should ponder the fallacy of the sorites which one can learn to live with. ...One cannot argue that because there is no plausible line to draw between a poor man and a rich man when it comes to single pennies that therefore there's no difference between the two of them. ...One cannot sensibly argue that because one cannot find a cut-off point therefore one must go on forever. And yet that's exactly the move that is made in connection with sentences: Because there is no plausible cut-off point (as indeed there seems not to be) therefore one goes on forever. No: One cannot sensibly go on forever in a natural language.<9>
I agree with Ziff's diagnosis of the above argument: it appears to be a pedestrian instance of a sorites fallacy. This is not to say that given the finite corpus of userintuitions about natural-languages a line is, after all, discernable. But the absence of a clear line is not an argument that sentences of arbitrary complexity are somehow actually available to us in some sense.
But if the argument above fails, then either there is some other argument that natural languages may be infinitely large, or else natural languages can't be infinitely large. What other argument is there that natural languages are infinitely large?
Perhaps consideration of the disquotatonal schema
"" means among English speakers that
along with the thought that the simplest syntactic theories for English will provide infinitely many syntactically well-formed English sentences lends support to the view that there will be infinitely many meaningful sentences of English and, by analogous considerations, of natural-languages generally. For many people seem willing to accept that each instance of the above disquotational schema will have to be true for instances where '' is replaced by a something that is a well-formed English sentence according to a correct syntactic theory for English.
But such an argument rests on a hasty generalization. For, even if there were reason to think that English syntax somehow provided for infinitely many sentences, there is no reason to suppose the disquotational schema applies to all of these. I doubt that there will be any such thing as a unique correct syntactic theory for English - there will always be rival but adequate syntactic theories for any language given finite sets of linguistic 'intuitions' as data. Perhaps psychological data will help, somehow, to pick out a class of syntactic theories that in some sense describe aspects of what English speakersrepresent to themselves in processing their sentences. But even if some such data isolated a single syntactic theory as the one that best describes such aspects of English speakers, and even if such a theory predicted that there were infinitely many syntactically well-formed English sentences, there is still absolutely no reason to believe that all of those sentences are meaningful for English speakers. To argue that all such sentences are meaningful it is necessary to provide some sort of characterization of meaning such that given such a syntactic theory the sentences it entails as well-formed are meaningful. That is, the present argument assumes that a theory of meaning has already been provided that makes sense of the idea that unusable sentences can be meaningful. I have no fancy a priori reasons for thinking that there cannot be such a theory. But no such theory seems forthcoming, as I have already argued, and it would run against the grain of the rather intuitive idea that expression-meaning is a matter of what can be used to mean by speakers.
I know of and can think of no other arguments that lends support to the thesis that natural languages are infinitely large. I also see no theoretical use that this supposition can have. So, I recommend that the supposition be dropped. To drop the thesis, however, is not to suppose that it is false. Perhaps it will turn out that there are unusable but meaningful sentences. My recommendation is that theorizing proceed on the basis of consideration of the usable. If we fail to come up with a theory of meaning for usable sentences, it seems unlikely that we will be able to come up with a theory of meaning that included the unusable ones.
Free from the constraint that provision has to be made for unusable but meaningfulsentences, the theorist of meaning can generalize over the sentences that are, in fact, meaningful for members of a community in the sense that they can be used by these people to say things to each other. If languages are infinite, then a theory that has it that the sentences of a language mean by virtue of facts about how they are used in such-and-such a way under these-and-those conditions will be questionable to the extent that it is implausible that anything interesting can be said about what people do or would do with unprocessable sentences. But without this constraining infinitude, a theory that made meaning a matter of correlations of sentences with propositional attitudes seems more of a possibility. I don't know what I would do if someone uttered a sentence that was more than two thousand words long with lots of center-embeddings or other complicated syntactic structures. But I am pretty sure that if someone uttered "Sinatra wants to tour next year with Metallica", I would come to believe that the person said that Sinatra wants to tour next year with Metallica, even if I never heard that sentence before. And I think that this is more important a fact for understanding the nature of meaning than the fact that the nicest syntactic theories that people have come up with for describing natural languages have allowed for infinitudes of sentences.
7.4 The Non-conventional and the Unuttered
Suppose that between you and I the expression "Grrr" has no conventional meaning but that either of us can use it to say that I am angry (ignore the indexical "I" here and throughout). This sort of situation can be made into what appears to be a counterexample to Schiffer's theory discussed above in chapter 6. For, consider thelanguage consisting of the single pair <"Grrr", that I am angry>. Call this language LG. And suppose that I utter "Grrr" on a certain occasion to say to you that I am angry. Suppose that the first time I utter "Grrr" to you was the first time that either of us had done so to indicate our anger and let's say that my reason for believing that "Grrr" would have its intended effect on you is RG. Whatever RG is, since "Grrr" has never been used before by either of us, it will not be about the past usage of "Grrr". Suppose next that we come to utter "Grrr" frequently to each other after this, but that each time we have completely forgotten our past uses. Each time we use it our reason for believing that "Grrr" will be effective is RG. This constitutes a practice among us of meaning in LG since often when one of us wants to mean that we're angry, we choose uttering "Grrr" to do so. Since both of us will come to have the appropriate M-sentence in our belief-box and since this M-sentence will have to contain as a part an M-sentence that means in M what "Grrr" means in LG, it follows that each of us must have an LG-determining translator since an LG-determining translator is by definition just something that produces such translations, whatever else it is. Thus, there is a practice among us of meaning in LG and processing of LG-sentences proceeds via an LG-determining translator. So Schiffer's theory predicts that LG is a language that we use, or, equivalently, that "Grrr" means among us that I am angry. But, it seems "Grrr" doesn't conventionally mean anything among us. Our reason for using "Grrr" is always RG and if the presence of this reason for using "Grrr" the first time one of us used it wasn't sufficient somehow for "Grrr"'s being meaningful, then surely it is still not sufficient for its being meaningful after "Grrr" has been often repeated. If "Grrr" wasn't meaningful the first time it wasused, then it won't be meaningful the n-th time it is used no matter how big n is. So, this sort of case seems like it is a counterexample to Schiffer's theory.
But the reason I didn't raise this sort of case in chapter 6 when I discussed Schiffer's theory is that I don't think it needs to be seen as a counterexample. I think that, given the conditions of the story, "Grrr" can be said to mean among us that I am angry, even if it doesn't conventionally mean among us that I am angry. My view is that the notion of conventional meaning need not be seen as what we should primarily be aiming for when we search for a theory of meaning. Meaning can be looked at as a more general phenomenon than conventional meaning.<10> I will call a case like the "Grrr" case a case of non-conventional expression-meaning. I think that no harm is done in allowing a theory of expression-meaning to capture both conventional and non-conventional cases of expression meaning.
My main reason for holding this is that I cannot think of any goals that the theory of expression-meaning could serve that would not be just as well served by the more general theory as by the narrower one that restricted itself to conventional meaning, except for the goal of marking the line between the conventionally meaningful and the non-conventionally meaningful. This latter goal will itself be partially served by the more general theory since all cases of conventional meaning will be cases of meaning more generally construed: meaning, though not a sufficient condition, is a necessary condition for conventional meaning.
The main goal of a theory of expression-meaning is to say what it is about our marks and sounds, etc. that makes them mean what they mean. It seems perfectlyreasonable to suppose that the important answer to this question will include "Grrr"-like cases. For, it seems perfectly reasonable to suppose that the important aspect of our marks, sounds, etc. with respect to their meaning is their ability to be used to say things. Conventional signs might be tools more ready to hand, requiring less creativity for their availability than non-conventional signs, but they mean what they do for exactly the same reasons that non-conventional signs mean, namely, they both can be used by one person to say something to another. I believe that this is the essence of the notion of meaning and that the distinction between conventional and non-conventional meaning is not so important.
It might also very well be that there will not be a principled distinction between the non-conventional and the conventional cases of meaning. Perhaps the two types of cases bleed gradually into one another and there is not a sharp line between them. The so-called arbitrariness of an expression - its arbitrary relation to what it expresses - has often been taken to be a mark of its conventionality.<11> But notice that both non-conventional and conventional expressions can show a high degree of arbitrariness. Consider June's buffalo dance discussed above in section 1.2.4. The dance was used by June to say to her friend Bingo that a buffalo is charging nearby. But the dance is not tied to buffalo charging by much more than sheer coincidence - it is not in any interesting sense a natural sign of nearby buffalo charging. But what needs to be added to arbitrariness and the ability to be used to say something to make an expression a conventional one? I have already argued in section 5.4 that the notion of convention itself on closer scrutiny seems too strong a notion to capture the conventionality of anexpression. I don't know if any weaker notions will be able to mark the distinction between the non-conventional and the conventional in a principled way. But my point here is that even if this distinction couldn't be marked neatly, we need not give up the aims of theorizing about meaning since these aims will be fulfilled by a general theory that accounts for non-conventional as well as conventional meaning.
Of course, there will be many unuttered utterance types that in the relevant sense could be used by one person to say something to another. I am willing to think that these also ought to be included in the sorts of thing a theory of meaning is supposed to be about. I don't see any evil consequence of slicing things up in this way.
7.5 The Theory of Meaning
As I am conceiving things, the important thing for a theory of meaning is to state under what conditions an utterance type can be used by one person to say something to another. This is a somewhat different project from stating what are the necessary and sufficient conditions for an utterance type to conventionally mean what it does since some utterance types can be used to say things even though they don't have conventional meaning. But as I have tried to argue, I think that this shift in focus might be the right way to go. But still, this would be better shown if I could state what were the conditions under which an utterance type were meaningful in my new sense and if I could give an analysis of the notion of meaning that I think I am after. Here I can only give general considerations.
Since I am not constrained by the idea that the natural-languages that we actuallyspeak are infinite in magnitude, I can make a move that seems to have been unavailable to Schiffer in his theorizing. Schiffer's L-determining translators were not tied in any specific way to processing and this led to the possibility of the sort of counterexample I presented in chapter 6 where an L-determining translator is employed as a proper part of an L'-determining translator for an L'-speaker who does not speak L. If Schiffer could have tied the input and output of his L-determining translators to something identifiable in the noodles of language users, he might have avoided this sort of problem.
So it might be considered progress if an L-determining translator is defined as something that actually puts into the belief-box of a hearer an appropriate M-sentence when a public-language sentence is heard. So, we might want to say that an L-determining translator now is a device that doesn't merely determine a mapping, but is one that actually causes a certain sentence to end up in the belief box: if I hear you say "Sinatra really swings," then my English-determining translator should put in my belief-box an M-sentence that means that you said that Sinatra really swings.
This move can partly block the sort of counterexample I gave to Schiffer's theory since now a device that is a proper part of an L-determining translator will be irrelevant unless it also puts things in the belief-box. The counterexample seems defused.
The price here is the finiteness of natural-languages. No actual L-determining translator will be likely to determine an infinite language given the actual facts about our nervous systems and their limitations. But as I have argued, the finiteness of natural-languages is not at all a foregone conclusion.
The story wouldn't be over here anyway. For amending Schiffer's definition ofan L-determining translator in the way that I have suggested does not lead to an adequate theory of meaning, even for the finite languages that such a device would in some sense determine. The theory we would have would be something like this:
P uses L just in case processing of L-sentences proceeds in the members of P by means of an L-determining translator.
The requirement by this theory is obviously too strong. It need not be that every time I process an English sentence I come to believe that somebody said something in uttering it. For the above theory is equivalent more-or-less to the following:
P uses L just in case whenever a member of P processes an L-sentence uttered by a speaker S, that member of P comes to believe that S said , where =L().
But if this theory states too strong a condition, then what weaker condition will do? To answer this question is to state a condition on the input to an L-determining translator. For, the trick is to say what are the meaning-determining conditions under which a sentence gets passed to the input of an L-determining translator.
The form of a theory of meaning will then be something like this:
P uses L just in case whenever condition C* is true of a member of P who is presented with an L-sentence , P comes to believe that the speaker said L().
I am not yet sure how to state what C* is.