Chapter 1
The Theory of Meaning: General Introduction

1.1 Introduction

What is it about some of the marks, sounds, and gestures that people produce and attend to that makes these marks, sounds, and gestures mean what they do among the people who produce and attend to them?

It is hard to believe that an essential property of the sound-type "lumi on valkoista" is the property of meaning among the Finns that snow is white. It seems just as hard to believe that the sound-type "lumi on valkoista" contingently has the property of meaning what it does among the Finns, but that it does not have that property contingently on its having some other property describable in physical and/or psychological terms. If that could be so, then two possible worlds could be identical in every physical and psychological respect, but differ in that in one world "lumi on valkoista" means among the Finns that snow is white, but in the other it means among the Finns that it's good to floss. It seems to me that if the sound-type "lumi on valkoista" has the property of meaning among the Finns that snow is white, it has this property in virtue of having some other property the description of which would require some sort of talk about how the Finns might use that sound-type to effect each other's psychological states. It seems, I think, inviting to suppose somewhat along the following lines: the sound-type "lumi on valkoista" means among the Finns that snow is white rather than that it's good to floss because it has the property of being such that for any two Finns, for example, Päivi and Martti, Päivi in producing a sound of that type in the presence of Martti will generally expect that Martti will come to believe that Päivi said that snow is white, and not that she said that it's good to floss.<1>

I take this sort of description as pointing the way toward what a general account of the nature of linguistic meaning should be like. But there are lots of problems standing between this simple description and a general theory of linguistic meaning.

The present dissertation attempts to examine ways of trying to formulate a general account of meaning which are more-or-less in line with the simple kinds of thoughts about meaning I just glossed. The ways of attempting such a general formulation that I am interested in work within a certain general conception about the nature of meaning. On this conception meaning can be understood wholly in terms of propositional-attitudes notions<2>.

The conception that I have in mind is due in the first instance to H. P. Grice<3> , but it is importantly elaborated by Stephen Schiffer<4> and David Lewis<5> . Other philosophers as well, among them P. F. Strawson, Jerry Fodor, Brian Loar, and Jonathan Bennett, have all, at least at some time or another, taken the conception I allude to as at least a working assumption.<6>

The general strategy of this dissertation will be as follows. In this chapter I will introduce Grice's conception of a theory of meaning in a general and somewhat informal sort of way, discussing various aspects of Grice's seminal paper "Meaning". Chapter 2 contains a sketch of what I take to be important aspects of the history of theorizing about meaning in the twentieth century, a rather overlooked area today. In chapter 3 I will state in somewhat more detail one way of thinking - or talking, anyway - about a general theory of meaning that will be employed by theories of meaning that I will consider in later chapters. Chapter 4 will contain a discussion of compositional-semantic theories and their relation to theories of meaning. I will argue, among other things, that it cannot be supposed that a language must enjoy a compositional-semantic theory in order for the language to be used by a group of people. But, most importantly, I will discuss an important counterexample to the view that language-processing must proceed via a compositional-semantic theory of some sort. In chapter 5 the issue will be theories of meaning that employ the notion of convention. I will begin with a discussion of David Lewis's various accounts of meaning. I will find a number of faults with these. First, I will argue that Lewis's theories presuppose that for a group of people to speak a natural language there must be a compositional-semantic theory for that language. Second, there will be problems for Lewis's theory that attend to his use of the notion of convention. Third, I will discuss a crucial problem, originally discussed by Stephen Schiffer, for Lewis's theories. After discussing Lewis's theories I will briefly discuss a theory due to Brian Loar. Loar's theory uses the notion of convention, but seems to avoid the crucial problem that Schiffer raised for Lewis's theories. But I will point out that in addition to inheriting all the problems associated with using the notion of convention and to requiring that usable languages enjoy compositional-semantic theories, Loar's theory has other important problems. In chapter 6 I turn to a recent theory of meaning offered by Schiffer which at first seems hopeful since it avoids commitment to compositional semantics from the outset. But I will present a serious counterexample to this theory. I will try to argue that the counterexample presents problems that are so serious for the type of theory that Schiffer has presented that there is really not much hope of trying to provide an adequate theory much like Schiffer's. Finally in chapter 7 will present a number of suggestions concerning the path I think a theorist of meaning might best take, given the discussion in the preceding chapters.

In the present chapter, then, I will get things started with a discussion of Grice's classic paper "Meaning" in section 1.2, followed by a brief discussion of the general goals of a theory of meaning in section 1.3.

1.2 Grice's "Meaning"

There are four themes that I consider central to Grice's seminal 1957 paper "Meaning" and that I would like to focus on in this introductory discussion. Any general understanding of the Gricean conception of meaning will require some understanding of each of these four themes. But they are important to look at here since they will arise in what follows in important ways. Here I will give a brief inventory of these themes, and then I will discuss each in a section of its own.

(1) The leading theme in "Meaning" is the thought that the semantic properties of conventional expressions are to be understood in terms of propositional-attitude psychology. I will call the notion of an expression meaning something the notion of expression-meaning.

(2) A second theme is about the nature of the relations between notions of meaning and notions of propositional-attitude psychology: do semantic notions reduce to notions of propositional-attitude psychology, and if they do, can we come to know of this reduction through philosophical analysis as this has been traditionally conceived?

(3) Another theme concerns the need for and nature of a notion of a speaker meaning something on a particular occasion by an utterance to achieve the understanding, mentioned above, of expression-meaning in terms of notions of propositional-attitude psychology. I will call the notion of a speaker meaning something on a particular occasion by an utterance, the notion of speaker-meaning. The idea behind this theme, then, is that the notion of expression-meaning can be analyzed neatly in terms of a notion of speaker-meaning (along with other notions that need not be worried about right now), and that the notion of speaker-meaning can be neatly analyzed wholly in terms of notions of propositional-attitude psychology. The overall result will be an analysis of expression-meaning wholly in terms of notions of propositional-attitude psychology.

(4) And the fourth theme that I am interested in is that of the possibility that notions of meaning that are specifically associated with language and, more generally, communication - like the notions of expression-meaning and speaker-meaning just alluded to - are interestingly related to notions of meaning that do not have to do with language or communication at all: Grice called one such non-linguistic, non-communication-related notion of meaning natural meaning. This theme ties into interesting questions that have been raised in objection to the Gricean program which I will also mention.

I will now discuss in turn and in a little more detail some aspects of each of these themes and their appearance in Grice's "Meaning" and, in some cases, in others of his works.

1.2.1 The Reduction of Expression-Meaning

The most important idea in "Meaning" is, roughly stated, that the semantic properties of linguistic expressions<7> can be understood in terms of propositional-attitude notions. I would like to discuss two matters with respect to this idea. The first is the generality with which Grice held the idea that semantic notions are to be understood psychologically: he seems to suppose that all the important semantic properties that linguistic expressions have are to be understood in psychological terms. The second is that Grice gave very little by way of the details of the story of how the psychological reduction of semantic properties of expressions is supposed to go; his sketchy remarks in "Meaning" really amount to little more than the suggestion of a program.

1. The specific way that Grice thought that the understanding of semantic notions<8> could be accomplished was by way of a notion of speaker-meaning (which I will discuss separately in section 2.3 below). He thought that semantic notions could be analyzed in terms of the notion of speaker-meaning (and other notions unimportant to us just now) and that the notion of speaker-meaning could in turn be analyzed in terms of propositional-attitude notions. The over-all effect would be the analysis of semantic notions in terms of propositional-attitude notions.

Thus, toward the beginning of the paper he criticizes a certain theory about meaning - that given by C. L. Stevenson in his book Ethics and Language<9> - on the grounds that it

...ignores the fact that the meaning (in general) of a sign needs to be explained in terms of what users of the sign do (or should) mean by it on particular occasions; and so the latter notion...is in fact the fundamental one.<10>

Here he shows his commitment to the view that expression-meaning, can be understood in terms of speaker-meaning. When we couple with this his view that speaker-meaning is to be understood in terms of propositional-attitude notions, we have the view that expression-meaning is to be understood in terms of propositional-attitude notions.

But the idea being discussed now is that all semantic notions that apply to linguistic expressions can be understood in terms of propositional-attitude notions, not just the notion of expression-meaning. Grice suggests this more general claim in the following passage:

If we can elucidate the meaning of

"x meant_NN something (on a particular occasion)" and
"x meant_NN that so-and-so (on a particular occasion)"

and of

"A meant_NN something by x (on a particular occasion)" and
"A meant_NN by x that so-and-so (on a particular occasion),"

this might reasonably be expected to help us with

"x means_NN (timeless) something (that so-and-so),"
"A means_NN (timeless) by x something (that so-and-so),"

and with the explication of "means the same as," "understands," "entails," and so on.<11>

Grice seems, then, to hold that the notions of synonymy and entailment, notions that philosophers often seem to take to be the ken of logic and not of psychology, are really founded in the notion of speaker-meaning, and therefore, by way of analysis, in the notions of propositional-attitude psychology. But the important point now is that the "and so on" in the above passage indicates that Grice took all semantic notions to beunderstood as psychological notions.

2. When Grice arrives at an analysis of speaker-meaning in terms of propositional-attitude notions, he only vaguely suggests how the notion of the meaning of a sentence, what I have called expression-meaning but Grice refers to as the "meaning_NN (timeless)" of a sentence, might itself be analyzed in terms of speaker-meaning:

"x" means_NN (timeless) that so-and-so" might as a first shot be equated with some statement or disjunction of statements about what "people" (vague) intend (with qualifications about "recognition") to effect by x.<12>

This is rather obscurely stated. I take it that the "qualifications about 'recognition'" are simply the details of his analysis of speaker-meaning. So we can, I think, safely restate as follows Grice's suggestion about the analysis of the notion of expression-meaning:

"x" means_NN (timeless) that so-and-so" might as a first shot be equated with some statement or disjunction of statements about what "people" (vague) mean_NN by x.<13>

But Grice does not discuss in Grice (1957) any details of the sort of "statement" that he has in mind in this suggestion. And he doesn't mention at all how speaker-meaning can help with other semantic notions like synonymy, understanding, and entailment which, as I pointed out above, he says it can be expected to help with. So what we have in Grice (1957) is but the barest outline of a conception of a general account of meaning, a suggestion for a program. In later chapters I will be scrutinizing various ways of trying to give the details of how expression-meaning can be understood in terms of propositional-attitude notions.

1.2.2 Reduction?

Above I talked about the reduction of expression-meaning to propositional-attitude notions. In her recent book, Meaning and Mind, Anita Avramides argues that it is perhaps best not to understand the analyses of the Gricean program as reductive analyses as followers of Grice such as Schiffer and Loar have.<14> She suggests that the Gricean conception of meaning would still be interesting if the analyses offered in it were understood as what she calls reciprocal analyses. In a reciprocal analysis the notions on the right side of the analytic biconditional are not to be taken, as in a reductive analysis, as somehow either clearer than or, in some other way, more basic than other notions in the "circle" that includes the analyzed notion on the left side of the biconditional:

The concepts mentioned on either side of the analytic biconditional have to be thought of as on a par.... Neither set of concepts would be more basic than the other.<15>

A reciprocal analysis remains interesting by showing "that understanding [the analyzed concept] is to be gained only by discerning its place in a system of interrelated concepts."<16> Thus, circularity - so long as it is not too blatant - in a reciprocal analysis is not necessarily cause for alarm:

Circularity, then, is a problem for the analysis of meaning only if our aim was to exhibit logical priority among concepts. Discovery of a circularity need not force us to abandon the project of analysis; we need only modify our claims.
... The problem of circularity plagues only reductive analyses. [footnote: "Reciprocal analyses are not open to the charge of circularity; or, one might say, circularity in a reciprocal analysis is never vicious. Of course, even reciprocal analyses must beware of traveling in circles that are too small."] One cannot have succeeded in breaking up a concept into simpler or more basic components if those components require the original concept for their explication. If beliefs are attitudes towards sentences, the goal of achieving a new level of concepts must be abandoned, but thisneed not compel us to abandon analysis altogether. Rather, we must accept that our analysis is of another kind, namely, reciprocal. What the analysis shows is how in precise detail our psychological and semantic concepts fit together. Here the analysis is like "a closed curve in space."<17>

Avramides also claims that "it is not altogether clear which interpretation [i.e. reductive or reciprocal] Grice himself intended of his work."<18> She quotes part of a passage from a later work of Grice to suggest that at least later on in his thinking he was not friendly to the reductive interpretation of his work.<19>

I will now briefly assess these views of Avramides concerning Grice's understanding of his own work.

Grice nowhere in "Meaning" uses the term 'reduction' (or some cognate). Furthermore, Grice nowhere in "Meaning" uses the connectives "if and only if" or "just in case" or some equivalent. Indeed, he doesn't seem to be consistent with respect the sort of glosses he is trying to supply in the article. Here is how he states a proposal for understanding the notion of an utterance meaning something on a particular occasion:

..."x meant_NN something" would be true if x was intended by its utterer to induce a belief in some "audience" and...to say what the belief was would be to say what x meant_NN.<20>

This, if true, states a sufficient condition for an expression to mean something on a particular occasion of utterance. After considering a counterexample to this proposal Grice seems, in the very same paragraph, to switch to concern with necessary conditions for an expression to mean something on a particular occasion of utterance:

...for x to have meant_NN anything, not merely must it have been "uttered" with the intention of inducing a certain belief but also the utterer must have intended an "audience" to recognize the intention behind the utterance.<21>

Again, later Grice seems concerned with necessary conditions:

Perhaps we may sum up what is necessary for A to mean something by x as follows. ...<22>

But consider the connectives used when Grice summarizes his thoughts in the paper:

"A meant_NN something by x" is (roughly) equivalent to....

"x meant_NN something" is (roughly) equivalent to....

"x means_NN (timeless) that so-and-so might as a first shot be equated with....<23>

Why is there this apparent inconsistency in Grice's discussion? Is he simply speaking loosely?

Answering these questions, I think, is complicated by reading Grice and Strawson (1956), "In Defense of a Dogma". This paper, famously, tries to defend against Quine's attacks something like the legitimacy of the family of expressions that include "analytic", "self-contradictory", "necessary", "synonymous", and "semantic rule".<24> In this paper, Grice and Strawson seem to be very clear about what form the statement of a reductive analysis should take. In talking about what Quine would allow as making satisfactory sense of one of these notions<25>, they say:

To make "satisfactory sense" of one of these expressions would seem to involve two things. (1) It would seem to involve providing an explanation which does not incorporate any expression belonging to the family-circle. (2) It would seem that the explanation provided must be of the same general character as those rejected [in Quine (1951)] explanations which do incorporate members of the family-circle (i.e., it must specify some feature common and peculiar to all cases to which, for example, the word "analytic" is to be applied; it must have the same general form as an explanation beginning, "a statement is analytic if and only if ...").<26>

A paragraph later, we are told the following:

It would seem fairly clearly unreasonable to insist in general that the availability of a satisfactory explanation in the sense sketched above is a necessary condition of an expression's making sense. It is perhaps dubious whether any such explanations can ever be given. (The hope that they can be is, or was, the hope of reductive analysis in general.)<27>

It would seem that if Grice in 1956 is this clear about the form that a reductive analysis has, then in 1957, if he intended such a thing, he could have been a lot clearer. Furthermore, the passage just quoted suggests that Grice in 1956 took the project of providing reductive analyses of notions to be a dubious one. So it may be tempting to suppose that Grice in "Meaning" didn't want to provide reductive analyses of the semantic notions he discusses: whatever he was up to, it wasn't the attempt to provide reductive analyses.

So, so far it looks like Avramides might be correct in saying that it is not clear what Grice's original sense of his own work was. But consider now the passage quoted by Avramides to suggest that Grice in his later work was perhaps against the reductive understanding of his program:

As I thread my way unsteadily along the tortuous mountain path which is supposed to lead, in the long distance, to the City of Eternal Truth, I find myself beset by a multitude of demons and perilous places, bearing names like Extensionalism, Nominalism, Positivism, Naturalism, Mechanism, Phenomenalism, Reductionism, Physicalism, Materialism, Empiricism, Scepticism, and Functional-ism.... After a more tolerant (permissive) middle age, I have come to entertain strong opposition to all of them, perhaps partly as a result of the strong connection between a number of them and the philosophical technologies which used to appeal to me a good deal more than they do now.<28>

Though this passage does suggest that he, when writing it, was against some sort of reductionism, it also strongly suggests that earlier in his life he supported it. Indeed, in his posthumous "Retrospective Epilogue" dated a year after the publication of the abovepassage, Grice answers in the following way J. Jack who Grice reports as having "reproved" him for attempting reductive analyses of semantic notions:

But what kind of analysis is to be provided? What I think we cannot agree to allow her [J. Jack] to do is to pursue the goal of giving a lax reductive analysis of meaning, that is, a reductive analysis which is unhampered by the constraints which characteristically attach to reductive analysis, like the avoidance of circularity.... ((In this connection I should perhaps observe that though my earlier endeavors in the theory of meaning were attempts to provide a reductive analysis, I have never (I think) espoused reductionism, which to my mind involves the idea that semantic concepts are unsatisfactory or even unintelligible, unless they can be provided with interpretations in terms of some predetermined, privileged, and favored array of concepts; in this sense of "reductionism" a felt ad hoc need for reductive analysis does not have to rest on a reductionist foundation. Reductive analysis might be called for to get away from unclarity not to get to some predesignated clarifiers.))<29>

This passage explains the senses in which Grice saw himself for and against the reduction of semantic notions. Clearly the sense in which he is for reduction shows that he would not be satisfied with an Avramidesean reciprocal analysis. Circles are not to be tolerated.

The concern to avoid circularity in the analysis of meaning is found in Grice (1957) when Grice twice remarks on circularities in Stevenson's theory.<30> Thus, Grice in "Meaning" can be understood as being concerned with finding a reductive analysis of semantic notions. He nowhere states or suggests that his motive for trying to find analyses of semantic notions is somehow to legitimate notions that might otherwise be suspect. So, we can take his word for it when he says that he has never supported reductionism in the sense of the above quotation. But this will not interfere with our supposing that Grice took himself to be trying to provide reductive analyses of semantic notions.

Of course, perhaps it still puzzling why, if reductive analysis was what Grice was after in 1957, he shouldn't have more clearly stated so given the clarity of Grice and Strawson (1956). A partial answer to this puzzle might be found in the fact that, actually, "Meaning" was not written after Grice and Strawson (1956), but long before it for a seminar he and Strawson gave in 1948.<31> Grice, it might be supposed, was not that interested in methodological subtlety in his seminar presentation and didn't dwell on being clearer on the points I am discussing in this section.

So far I have only argued for the extremely modest point that Grice, anyway, took himself to be trying to find reductive analyses for semantic notions. The question still remains whether Avramides is correct in thinking that the search for such analyses is misguided somehow. I cannot argue the point here, but I will merely assert that I do not think that Avramides is correct and that it is perfectly reasonable to seek reductive analyses of at least some important notions, meaning among these.

1.2.3 Speaker-Meaning

The attempt to explicate a notion speaker-meaning plays a central role in Grice's "Meaning". Grice's leading idea was, again, that expression-meaning should be explicated in terms of some notion of speaker-meaning and that speaker-meaning should be explicated in terms of notions of propositional-attitude psychology. In this section I will talk first a bit about Grice's analysis of speaker-meaning and then about the status of this notion within the theory of meaning as a whole.

Roughly<32>, Grice's analysis of speaker meaning has it that a speaker S speaker-means a proposition p by an utterance u just in case S produces u with the threeintentions (a) of getting an audience A to believe p, (b) of having A recognize that S's intention is that A believe p, and (c) of it being part, at least, of A's reason for believing p that A recognized S's intention to get A to believe p. It is common for people first encountering a formulation of Grice's account of speaker-meaning to be baffled by its apparent complexity. But, the complexity is well motivated, and Grice's grasping the motivations here should probably be considered one of the genuine examples of innovation to be found in twentieth-century philosophy.

I will discuss below to what extent the notion of speaker-meaning is related to any ordinary notions. But a good mnemonic is that the ordinary notions of saying or telling as, for example, in the sentences [a] and [b] come somewhat close to speaker-meaning:

[a] Digby said that the game was exciting.

[b] Gerti told me that we would be on time.

The notions of saying or telling are both notions of communication. To say or to tell something to someone is to communicate something to them. And it is helpful to keep the ordinary notion of communication, or something like it, in mind in discussions of speaker-meaning. In the following informal discussion I will use saying, telling, communicating, meaning, and speaker-meaning somewhat interchangeably.

The first thing to notice about speaker-meaning is that it requires that the speaker intends to get an audience to have a certain belief<33> . But the second thing to notice is that not just any case of getting a person to have a belief is a case of a speaker meaning (saying or telling) something to somebody. If an action of mine isto be properly called an act of telling my guest that I am out of coffee, I must intend that my guest come to believe that I am out of coffee. But the action with this intention by itself doesn't constitute telling. For I can perform an action with the intention that my guest comes to believe that we are out of coffee which is clearly not a case of telling. For example, I can just show my guest that my coffee canister is empty with the intention of their coming to believe that I am out of coffee. So performing an action intending that an audience come to believe something is necessary but insufficient for speaker-meaning.

If I perform an action with the intent that you come to believe something, I am presumably going to present you with some sort of evidence for the belief. Showing you the empty coffee canister is evidence for the belief that I am out of coffee. The thing that Grice seems to be the first to notice about all this is that in cases where we are trying to communicate, to say or to tell, something to some audience, the evidence that we present the audience should not be evidence that would have gotten the audience to have come to have the appropriate belief if it somehow were available to them apart from being shown by somebody. If my guest happened to see my empty coffee canister without my showing it, this would have amounted, presumably, to the same evidence for them that I was out of coffee as did my presenting it.

Consider, to use an example drawn from Grice, the difference between showing someone a photograph and presenting them with a drawing. A photograph will generally be evidence that what it depicts occurred, and this will be so independently of whether the photograph was intended to convey something to somebody. But a drawing depicting an occurrence does not - by itself - count as evidence that what it depicts occurred. If you show me a photograph of a situation, I will, in the usual case, come to believe that that situation occurred, but my conclusion that it occurred will be drawn on the basis of what the photograph depicted.<34> But if you show me a drawing of a situation, since the drawing itself cannot be considered evidence that the situation occurred, if I am to come to the conclusion that the situation depicted, in fact, did occur, I will have to use as evidence for this conclusion not just what the picture depicts, but, presumably, that I took it to be your intention that I should believe what the drawing showed. That is, if I thought that you were showing me the drawing in order to get me to believe that what it depicted was the case, then I may well conclude - supposing I deem you trustworthy - that what it depicted was, in fact, the case. So, the fact that you intend to get me to believe that some state of affairs holds can count as evidence for me that that state of affairs, indeed, does hold.

Grice saw that this sort of evidence on the basis of recognition of intention is at the heart of communication, of speaker-meaning. If a speaker is to tell someone that some state of affairs holds, the speaker must perform an action intending to get an audience to believe that that state of affairs holds on the basis of their recognizing the speaker's intention to get them to so believe. The speaker's intention to get the audience to believe that the state of affairs holds itself must be part of the audience's reason for coming to believe that the state of affairs holds. Or, at least, the speaker should intend it to be so.

There are important counterexamples to both the necessity and the sufficiency of Grice's analysis of speaker-meaning<35>, but it is not my purpose here to give a comprehensive report on this aspect of Grice's view of meaning. The details of Grice's analysis of speaker-meaning won't play a big role in the discussion in subsequent chapters. But it will play some role, and ultimately, in chapter 7, I will suggest an a variation to Grice's analysis. But there's no need to anticipate the details of all that here.

I turn now to the discussion of the role of the notion of speaker-meaning in the theory of expression-meaning.

It is not altogether obvious that there is a pretheoretically clear notion of speaker-meaning to be found and analyzed. This point has been at least hinted at by a number of writers.<36>

Grice himself in "Meaning" at one point turns our attention away from the schema

A meant_NN by x that so-and-so (on a particular occasion)

which he begins with to the notion of telling:

What we want to find is the difference between, for example, "deliberately and openly letting someone know" and "telling" and between "getting someone to think" and "telling."<37>

This move seems to indicate some sort of unclarity in just what notion Grice was after. In Grice (1969) Grice speaks of "the notion of saying that p (in the favored sense of say)".<38> And in Grice (1968) he speaks of "what the speaker has said (in a certain favored and maybe in some degree artificial, sense of 'said')".<39>

So, though the basic formula of the Gricean program - something like, "expression-meaning to speaker-meaning and speaker-meaning to speaker attitudes" - may seem attractive enough at first glance, there really is quite a bit that is vague here. And I don't think that the vagueness can or need be finally or completely dispelled in advance of the consideration of proposals for actual theories of meaning. But, still, I will try to clarify things at least a little on this matter.

What is wanted is a reductive analysis of the notion of a sentence's meaning what it does among a group of people, that is, of the notion of expression-meaning. One thing that this means is that concepts used in the desired analysis must be themselves analyzable without recourse to the notion of expression-meaning: the analysis of expression-meaning shouldn't be circular.

One way of proceeding in this endeavor will be to try to analyze expression-meaning in terms of propositional-attitude notions and one or more pre-theoretically available propositional speech-act notions like saying, telling, asking, commanding, etc. (along with perhaps non-semantic and non-psychological ancillary notions that need not concern us here). To succeed in such an endeavor would be quite an achievement, in my view, even without analyses of whatever speech-act notions were employed. But, to complete the entire project, the analysis of the speech-act notions would have to be achieved in terms that didn't require the notion of expression-meaning. I think that this overall strategy is probably the best and it informs much of what I say in the last chapter of this dissertation. But it is not the only strategy.

David Lewis, more-or-less, offers an analysis of expression-meaning without using, or proceeding through the use of, any propositional-speech-act notions. Lewis at one point does try to suggest that, after all, his theory will entail that expression-meaning is a sort of regularity of speaker-meaning among a group of people.<40> But, even if Lewis were wrong about this,<41> I should think that it wouldn't matter to the project of trying to understand the nature of expression-meaning. I see no good reason why an analysis ofexpression-meaning should have to say or entail anything about a notion of speaker-meaning to be legitimate. I doubt that a serious theory of expression-meaning can avoid doing so - that is in part why I think the best strategy is to make free use of propositional-speech-act notions in the analysis of expression-meaning -, but I don't pretend to know of any a priori reasons why it should not avoid doing so.

It seems to me that through his life, Grice's own strategy for attempting to elucidate expression-meaning was a mixture of the two that I have just discussed. Instead of attempting to stick with antecedently available speech-act notions, he attempted to isolate a speech-act notion that would ultimately allow for the elucidation of expression-meaning. But even if his attempt to isolate such a notion was not an attempt to clarify an antecedently available speech-act notion, neither was it an attempt to construct a wholly artificial notion. Recall the passage quoted above at the beginning of section 2.1 in which Grice complains of causal theories that they fail to help elucidate the notion of a speaker meaning something on a particular occasion. Whatever the notion of speaker-meaning Grice wanted was, however, the ultimate interest of it is absorbed in the question of how successful a theory of expression-meaning that employs it might be.

1.2.4 Natural and Non-Natural Meaning

"Meaning" begins with the suggestion of a distinction between what Grice calls natural meaning and what he calls non-natural meaning. As examples of cases of natural meaning Grice gives:

Those spots mean (meant) measles.
The recent budget means that we shall have a hard year.<42>

As examples of cases of non-natural meaning, Grice gives:

Those three rings on the bell (of the bus) mean that the bus is full.

That remark, 'Smith couldn't get on without his trouble and strife,' meant that Smith found his wife indispensable.<43>

The natural/non-natural meaning distinction is indicated by Grice by giving the above examples and noting five differences in the kinds of inferences that can be drawn in the various cases. For example, he notes that it seems odd to say "those spots meant measles, but he hadn't got measles", but okay to say "those three rings on the bell (of the bus) mean that the bus is full, but it isn't in fact full - the conductor has made a mistake".<44> Grice also tells us that he would like cases of speaker-meaning as well as of expression-meaning to be counted as cases of non-natural meaning<45>. And cases in which the word "mean" (or cognates) are used in the sense of purpose or intend - as in "John meant to do the laundry" - Grice indicates he would like to include as cases of natural meaning. Classifying these various senses of "meaning" as Grice suggests seems right enough given the remarks by Grice which were alluded to just above about how to draw the natural/non-natural meaning distinction. But Grice does not really provide a sharp demarcating criterion between natural and non-natural meaning. In fact, he leaves off entirely the discussion of natural meaning to focus on the analysis of non-natural meaning after the following remarks:

The question which now arises is this: "What more can be said about the distinction between the cases where we should say that the word is applied in a natural sense and the cases where we should say that the word is applied in a nonnatural sense?" Asking this question will not of course prohibit us from trying to give an explanation of "meaning_NN" in terms of one or another natural sense of "mean."
The question about the distinction between natural and non-natural meaning is, I think, what people are getting at when they display interest in a distinction between "natural" and "conventional" signs. But I think my formulation is better. For some things which can mean_NN something are not signs (e.g. words are not), and some are not conventional in any ordinary sense (e.g. certain gestures); while some things which mean naturally are not signs of what they mean (cf. the recent budget example).<46>

Grice is presenting here an aspect of his thought that links it up to a long tradition of thinking about language that begins at least as far back as Plato's Cratylus where a distinction between conventional and natural meaning was drawn perhaps for the first time.<47> My discussion right now will focus on the question of what Grice might have meant when he spoke of an "explanation of 'meaning_NN' in terms of one or another natural sense of 'mean'".

One very general thing that Grice might have meant by speaking of an "explanation" of non-natural meaning in terms of some sort of natural meaning is that, ultimately, one should expect that expression-meaning and speaker-meaning will be analyzable in terms of some sorts of non-intentional notions. The idea, presumably, would be that, first, there would be the two-tiered reduction of expression-meaning - expression-meaning would be explained in terms of speaker-meaning and speaker-meaning would be explained in terms of propositional-attitude notions -, but, then, propositional-attitude notions would be explained in a certain way by wholly non-intentional notions. The explanation of propositional-attitudes in non-intentional terms would have to talk about the relations that held between organisms and the world such that organisms could have intentional states about the world. These organism-world relations would amount to some sorts of relations of natural-meaning. So, the idea would be, that Grice, when he speaks about the "explanation" of non-natural meaning in terms of natural meaning, has in mind that ultimately the analysis of public-language semantic notions will have to lead us to some sort of non-intentional explication of propositional-attitude notions and this explication will use some notion or other of natural-meaning.<48>

If Grice meant this by his comment, then he was simply indicating that things might be more-or-less like what some of the theories I will discuss in the later chapters of this dissertation make them out to be. And I don't think it is altogether odd to read Grice's comment this way. But in fact, I think he had something else in mind with the comment which I will now discuss a bit.

For Grice, we might say, all language is founded in speech somehow, but speech is at least conceptually possible without language. This is just to summarize the idea that expression-meaning is reducible to speaker-meaning (along with some other notions that are not important here): if speaker-meaning somehow turned out to require expression-meaning, then whatever the relation expression-meaning stood in to speaker-meaning, it wouldn't be one of reduction. But speech is essentially the production of some behavior or artifact to communicate something to somebody. If speech is possible without language, then some behaviors or artifacts will be capable of being used in communication even though they don't have conventional meaning. But since such behaviors or artifacts are capable of being used in communication, that is, capable of being used by one person to get another person to have a belief, we might just be tempted to suppose that these behaviors or artifacts have meaning in some sense, even if this sense is not that of expression-meaning. I believe that this sort of reasoning was behind the traditional notion of natural meaning and that to some degree it lies behind Grice's interest in trying to clarify such a notion as well as his suggestion that there might be something of an "explanation" of non-natural meaning in terms of natural meaning. If all language is conceptually traceable somehow back to language-less speech and language-less speech requires behaviors or artifacts that naturally mean, then all language has been traced to natural meaning. I think that this is the sense in which Grice might have felt there might be an "explanation" of non-natural meaning in terms of some notion of natural meaning.<49>

I think there is both something right and something wrong in this line of reasoning. There are two parts to what I think is right here. (1) If the Gricean picture of language is correct, we do somehow expect that every bit of language must somehow be traceable - conceptually, not historically - to non-linguistic speech. And (2) a behavior or artifact that is used in non-linguistic speech must have some feature that makes it suitable for its use in communication. What I think is wrong here is that the feature of a behavior or an artifact used in non-linguistic speech that makes it suitable for communication needs to somehow exploit behaviors or items that naturally mean something.

It seems to me that what Grice probably has in mind is a case like the following. Suppose that June and Bingo, two language-less people, have noticed that whenever the big shady tree shakes in a certain way a buffalo is charging nearby. So, we may suppose the shaking in that way of the big shady tree naturally means that a buffalo is charging nearby. One day June wants to tell Bingo that a buffalo is charging nearby and tries todo so by attempting to shake the big shady tree - which we may suppose was not shaking by itself on this occasion - in that certain way. Supposing that such a case might be a case of language-less telling, we can see that there is a sense in which natural meaning is exploited to achieve non-natural meaning, in this case, to tell somebody something.

But it isn't hard to imagine cases in which there is no real connection to anything that naturally means something in any obvious sense.<50> Imagine another scenario in which June wishes to get Bingo to believe that a buffalo is charging. Suppose that June and Bingo have known each other a long time and that once, some time before the present occasion, June and Bingo saw a charging buffalo and June suddenly began to dance playfully and laugh. Suppose also that June and Bingo have seen many buffalo without June's doing the buffalo dance and that a few times June did the buffalo dance for fun without a charging buffalo about. This last supposition, of course, is to ensure that June's buffalo dance is not like that certain shaking of the big shady tree which could be taken to naturally mean that a buffalo was charging nearby, I presume, because its shaking in that way was sufficient in June's and Bingo's experience to indicate that there was a buffalo charging nearby. Suppose further that just a week earlier June and Bingo were in a somewhat similar situation to the present one but in the earlier episode Bingo seemed to June too close to the path of a charging buffalo. On that occasion June did the buffalo dance but with a face that she knew Bingo would take to be incongruous with June's merely playing. June did this hoping that it might cause Bingo to suspect that a buffalo was about, look about for it, see the buffalo that was, in fact, charging, and move herself to a safer distance from its path. None of what June hoped for happened however. Bingo, it turned out, was at a safe enough distance when the buffalo past by. But upon seeing the buffalo pass by Bingo realized what June was hoping to do. And June came to believe that Bingo upon seeing the buffalo pass by realized what June hoped to achieve by doing the buffalo dance with a face she knew Bingo would take to be incongruous with play. So, now June wants to get Bingo to believe again that a buffalo is charging. She remembers all the episodes just mentioned and she believes that Bingo remembers them too. And she believes because of all of this that if she does the buffalo dance with a face that she knows that Bingo is likely to take as incongruous with play, Bingo will recognize that June intends to get her to believe that a buffalo is charging nearby and thereby, in fact, come to believe that a buffalo is charging nearby. And so she does the buffalo dance with an appropriate expression intending Bingo to reason as she has imagined.

June in this case, it would seem, has succeeded in telling Bingo that a buffalo is charging. But she has not exploited anything that could easily be understood to naturally mean anything. What made it possible for June to succeed in communicating with Bingo in this case might be described as an accidental feature of their shared history that could somehow be exploited to achieve communication. I don't know if there is anything interesting in common between this sort of case and cases like the first one I considered where it is remotely plausible that items that naturally meant were used to achieve communication. And I don't know if there are further cases that ought to be considered. My point is simply that exploitation of items that naturally mean doesn't seem to be a necessary feature of language-less speech.

Of course, there is the not-so-interesting common element in the two cases that I have already mentioned: there must be some feature of an artifact or a behavior used in language-less speech that makes it suitable for communication. One could define natural meaning as the having of such a feature, but it is hard to see what would be gained by doing so.

The sorts of issues that are involved here will arise again at the very end of this dissertation. And they are not unimportant to the Gricean. A common objection to Griceanism takes the form of the observation that it seems unlikely that many of the things that we say in our everyday talk could ever really be thought if we didn't have language, let alone expressed in language-less speech.<51> For example, it may seem hard to imagine that language-less creatures could have a thought like that if next Thursday it rains, perhaps it will be better not to try to travel to my cousin's house unless I can get a ride with a friend. And even if you could convince yourself that a language-less creature can have some such thought, still, it is hard to see how anything could be usable for expressing such a thought to another unless there were some conventional devices for doing so. So, the objection has it, the Gricean program is implausible because there seems to be no way that a sentence that has as its meaning a complex thought like the above could derive its meaning from non-linguistic communicative intentions.

Griceans have responded to this sort of complaint by sketching stories or suggesting sketches of stories that have it that things start off somewhat like in the June and Bingo stories above. A little language gets introduced. Then, once that language is introduced, this makes it possible for more complex thoughts to be available. For example, once June and Bingo have a conventional way of expressing the thought that a buffalo is charging, it is conceivable that they now have the ability to see the activities of other creatures as charging activities because these activities are similar enough to the charging buffalo cases. And perhaps by a clever variation of the conventional device developed for mentioning charging buffalos, bolting rhinos can be talked about as well. Then after repetition, this sort of communication becomes conventional and part of June and Bingo's shared language. And so on. A little thought, some communication, then language; then a little more thought, some creative extensions of already existing language in new communications, and then after repetition, new language; etc. And in this way thought precedes language, but also depends on it and eventually the richness of our own linguistic devices can be arrived at.<52>

The matters here are interesting and, as I say, not unimportant to the Gricean, but I will not be able to discuss them further in this dissertation.

1.3 Gricean Theories of Meaning

I will talk now briefly and in general terms about the goals of the theory of meaning and I will distinguish various sorts of theories that, I think, can all fairly be called Gricean. And I will indicate the sort of theory that the present dissertation will aim at.

In chapter 3 I will try to motivate a little better the view that we distort nothing essential to our purposes of coming up with a general theory of expression-meaning if we begin by considering the abstract case of languages without either indexicals or non-indicative moods. For now I will just state that that will be what I will do throughout most of this dissertation.

Supposing that abstracting in that way is appropriate, then, we can state a primary goal of theorizing about expression-meaning as, more-or-less<53>, the provision of a satisfactory completion of [M]<54>:

[M](P)(x)(y)(x means among the members of P y ...)

It is extremely difficult if not impossible to say ahead of actual critical work on particular theories what should count as a satisfactory completion of [M]. But someone working within a broadly Gricean framework can say a little bit about this anyway.

If we are after a Gricean theory of meaning in a pretty strong sense, then the completion of [M] must not contain any vocabulary that expresses public-language semantic notions or any vocabulary that expresses notions that require for their applicability the applicability of public-language semantic notions. To guarantee that this requirement has been met may be very difficult. For, Gricean theories of expression-meaning, in a slightly broader sense now, are generally aimed to provide glosses of semantic notions in terms of propositional-attitude notions. To guarantee that the condition under discussion has been met will be to guarantee that propositional-attitude notions themselves don't require for their applicability the applicability of any public-language semantic notions. I do believe that there is some reason to think that propositional-attitude notions won't presuppose public-language semantic notions, but I don't know of any conclusive arguments that this is so. But I can't discuss the matter further here.

I will be happy, however, in the discussion to follow to count as a satisfactory completion of [M] any completion that makes free use of propositional-attitude and propositional-speech act notions so long as none of the other notions it uses presuppose public-language semantic notions. I will count such a completion as a Gricean theory of meaning. But it may be useful to qualify this: I will call such a theory a weak Gricean theory. To have a strong Gricean theory of meaning, let's say, is to have the conjunction of a weak Gricean theory and a theory of propositional-attitude and propositional-speech act notions that shows that such notions don't themselves presuppose any public-language semantic notions. One might want to speak of a third sort of Gricean theory as well: we might say a moderate Gricean theory was one that glossed public-language expression-meaning in terms only of propositional-attitude notions along with some other non-public-language semantic notions, but that did not avail itself of propositional-speech act notions.<55> Grice, then, himself was apparently only aiming at a moderate Gricean theory in trying to elucidate the notion of speaker-meaning - a propositional-speech act notion - in terms of propositional-attitude notions. That is, he used a propositional-speech act notion as a stepping stone to a theory of expression-meaning that did not, in fact use such notions. Another way of putting this which uses the terminology I suggest here is that Grice used a weak Gricean theory to move on to a moderate Gricean theory.

This is clearly not at all a complete discussion of the constraints on satisfactory completions of [M]. But, as was said, it is probably the best that can be provided ahead of critical work on actual theories, at least for someone who is interested in trying out the Gricean program. It always remains an open possibility that some bit of vocabulary turns out, after critical assessment, to be less clear than it was thought to be.

A satisfactory completion of [M], then, would amount to a theory of meaning for languages with no indexicals or non-indicatives because it would describe in terms of propositional-attitude and propositional-speech act notions the conditions necessary and sufficient for a sentence to mean some specific thing among the members of a population.

1.4 Summary

In this chapter I introduced some important aspects of the Gricean program in the theory of meaning by discussing various aspects of Grice's article "Meaning" and I discussed in a general way the goals of a theory of meaning. In the next chapter I will give a brief history of the investigation of meaning in the twentieth century. Then in the subsequent chapters I will move on to particular issues involved in providing a theory of expression meaning.