NON-TRUTH-CONDITIONAL MEANING

There are two dominant approaches to semantics. One sees the task of semantics as to provide a systematic account of the truth conditions of (actual and potential) sentence uses. The other assumes that a use of a sentence expresses a statement (proposition, thought—terminology varies here), a statement being the sort of thing that can be asserted and believed, and also the sort of thing that, as a representation of how the world is, can be assessed as true or false. The task of semantics, on this view, is systematically to spell out how sentence uses are associated with statements.

While the aims of the two types of theories are different, they are related. A use of a sentence to make a statement is, after all, presumably true (or false) in virtue of the truth (or falsity) of the statement made. Hence, to assign statements to sentence uses is to assign those uses truth conditions. Thus both approaches give pride of place in semantics to an account of how sentence uses come to be true or false.

No one thinks that giving an account of truth conditions or of what statements say, for a language, says all there is to say about conventional meanings of expressions in the language, though exactly what more there might be is a matter of controversy. Here are the main candidates for what might be left out of such accounts.

Mood and Force

The theories just discussed aim to illuminate what is going on when one uses the sentences of a language to make assertions, to commit to the truth of a claim. But, of course, we can do much more than make assertions with our sentences, and some aspects of conventional meaning are obviously keyed to doing things other than asserting. Examples are grammatical and phonological forms associated with questioning, ordering, and exclaiming. It is a fact about conventional meaning if anything is, that subject/auxiliary inversion is used to question in French, German, and English, that prefixing a declarative sentence in English with "if only" signals a wish, that sentences such as "Yuck!" and "Damn it!" express attitudes that are not to be evaluated as true or false. One task not discharged by truth-conditional or statement semantics, then, is detailing when and how linguistic forms have as part of their conventional meaning the task of signaling that a particular sort of speech act (asserting, questioning, promising, warning, expressing disgust, etc.) is being performed.

One might question the extent to which this is more than just an appendicle to truth-conditional or statement semantics. One might say that interjections like "Grody!" and "Awesome!" are elliptical for truth bearers ("That is grotesque!" "That's awesome!") uttered with a particular force. Whether or not this is so, the interjections do not combine with connectives and the range of sentences in the language to produce complex sentences; their meanings, if different from that of declaratives, would thus seem to be walled off from other aspects of meaning. There seems to be a rather small catalog of devices, like auxiliary inversion and the subjunctive, to indicate force; such devices, furthermore, do not seem to be iterable, as constructions that contribute to truth conditions are. While one can disjoin a negation, then enclose the result inside the consequent of a conditional, etc., force indicators seem by and large to exclude one another (one cannot, for example, turn the optative "would that he were gone" into a question). Furthermore, it is not clear that any particularly novel sort of meaning is required in an account of the meanings of, for example, orders and questions. One might suspect that in some sense the content of the declarative "You will sit" and of the imperative "Sit!" are the same, the difference lying only in the force of their utterance. Perhaps questions have a slightly novel meaning. For example, it is often suggested that the meaning of "Who will sit?" is something like the set of (contextually relevant or possible) answers to it. But this makes the meaning of a question just a set of statements.

J. L. Austin once claimed that a good deal of natural-language vocabulary has meanings whose job is to signal that one is, and is only, performing a (nonassertive) speech act. For example, on Austin's view, to utter "I promise to meet you at 5:00" is not to assert anything, but to make a promise. Austin (1962) gives a lengthy catalog of verbs (part of) whose conventional meaning, he claims, is to signal (when used in the first-person present) that a particular speech act is being performed, representative examples being "acquit," "nominate," "bet," "toast," and "concede." He suggests that the number of such verbs contained in English is "of the third order of the power of 10."

There are arguably many expressions whose purpose is in part or in whole to signal that, whatever else the speaker might be doing, he is performing a particular nonassertive speech act, though exactly which expressions do this is a matter of controversy. "Just between you and me" (as in "Just between you and me, the provost hates the president") might be a conventional means to warn or ask one's audience not to divulge the information imparted by the rest of the sentence. Racial slurs are, inter alia, conventional means of insulting and displaying contempt for their targets, as are the merely obscene or insulting things we may call someone in the course of commenting on them. Presumably, though, to utter something like "That jerk Smith is at the door" is to say something true or false, depending (only) on whether Smith is at the door.

Conventional Implicature

Grice (1967/1989) drew a distinction between what the use of a sentence "strictly" says and what it implies. Both what is said and what is implicated are statements. Indeed, what a sentence use says, in Grice's sense, seems to be the statement that a semantic theory (of the second sort discussed above) aims to assign to the use. According to Grice, it is what a sentence strictly says, and only what it strictly says, that is relevant to the question of whether the use of the sentence is true.

What, then, is the role of what is implicated by the use of a sentence? Some such implication is a one-off affair, as when one says, "There's an umbrella in the closet," expecting one's auditor to work out that rain is in the offing. Implication of this sort exploits facts obvious to all—for example, that speakers generally try to say helpful and relevant things—to efficiently convey information; it allows us to convey much more than our words literally mean.

Grice distinguished this sort of implication—conversational implicature, as he called it—from cases in which "the conventional meaning of the words used … determine[s] what is implicated, besides helping to determine what is said" (Grice 1967/1989, p. 25). Grice's examples were the words "therefore" and "but." In uttering "A ; therefore B," Grice claimed, I say that A, say that B, commit myself to B 's following from A, but I have not "said (in the favored sense)" that B follows from A : "I do not want to say that my utterance … would be, strictly speaking, false should the consequence in question [fail] to hold" (Grice 1967/1989, pp. 25–26). In uttering "He is F but G," one speaks truly, Grice said, just in case the relevant individual is F and G, though one clearly conveys some sort of contrast between being F and being G. To use "therefore" or "but" is to commit to these implications. Since the implications are carried by the very words used, they are not one-off conversational implicatures but conventional implicatures.

A rather large class of expressions have been said to give rise to conventional implicatures. Karttunen and Peters (1979) suggest that words and constructions often said to give rise to presuppositions in fact give rise to conventional implicature. Here are some examples, with the word purportedly carrying the conventional implicature italicized and the implicature roughly indicated in parentheses:

Even John understands it. (John is unlikely to understand it.)

Martin still loves her. (Martin loved her in the past.)

Jed failed to pass. (Jed tried to pass.)

Other examples of purported conventional implicatures are nonrestrictive relative clauses and appositives. "Martina, a yogi, hunts bears" commits the speaker to Martina's being a yogi, but arguably would be true even if she is not one, so long as she does hunt bears.

It is controversial whether there is such a thing as conventional implicature. Bach (1999) argued that a complete report of Bob's utterance of "Even Mo likes Jo" is given with "Bob said that even Mo likes Jo"; simply saying, "Bob said that Mo likes Jo" is not giving a complete report. Since "that even Mo likes Jo" is here specifying what Bob said, Bach concluded, part of what Bob's utterance says must be (something like the claim) that Mo's liking Jo is unexpected. But if that is part of what is said, then the utterance is true only if it is unexpected that Mo likes Jo. According to Bach, this sort of argument shows that pretty much every expression alleged to carry a conventional implicature in fact does not.

It is not clear that this argument succeeds in showing that conventional implicatures are a fiction. "What is said (by utterance u )," as used by Grice, is a technical term. The phrase and its cousins have an everyday use as well. It is not at all clear that Grice assumed that if an utterance would naturally and correctly be reported as saying that p, then p must be part of what it says in the technical sense. We are, after all, pretty loose in how we report indirect speech.

One might hold that conventional implicatures are just as much said by a use of a sentence as anything, but have properties and relations to sentence uses that make it worthwhile to distinguish them from other claims literally made by sentence uses. Christopher Potts (2005) distinguished what he called "at issue" claims made by a sentence use (roughly, what Grice had in mind by "what is said") from conventional implicatures. (However, Potts's view, unlike Grice's, is apparently that conventional implicatures are relevant to truth conditions. He takes conventional implicatures to be "entailments," and holds that sentences carrying such implicatures can typically be paraphrased by conjunctions, one conjunct of which is the implicature.)

For Potts, one putative difference between conventional implicature and at-issue content is that even when a speaker embeds an expression carrying a conventional implicature, the speaker becomes committed to the implicature; this is not so with at-issue content. To see the point, consider "Bob, a linguist, likes clams," where the at-issue content is that Bob likes clams and the conventional implicature is that Bob is a linguist. When one embeds the sentence under negation or an attitude verb (as in "It is false that Bob, a linguist, likes clams," or "Mary said that Bob, a linguist, likes clams"), use of the resulting sentence seems to commit the user to the conventional implicature, but not to the at-issue content.

A conventional implication is like a presupposition in this regard. Potts argues that conventional implicatures are not presuppositions, since false conventional implicatures and false presuppositions have different effects. When a presupposition of a sentence is false, the assertion of its at-issue content is unfelicitous, perhaps without truth value; this is not so with conventional implicature. In the case of conventional implicature, that Bob is not a linguist does not impugn or cast doubt on the claim that Bob likes clams. Knowledge that the presupposition of "It was Bob who stole the book (namely, that someone stole it) is false makes the assertion that it was Bob who did it unacceptable.

Nonpropositional Meaning

What is conventionally implicated has truth conditions. A non-truth-conditional conventional implicature does not enter into the truth conditions of the use of a sentence; its truth or falsity is not relevant to the truth or falsity of the sentence use implicating it. Other alleged sorts of non-truth-conditional meanings, however, are non-truth-conditional in the sense that they simply are not the sort of thing that can be true or false—they are, as it is sometimes said, not truth-apt.

One (alleged) example of such a meaning is presented by those who hold that linguistic meaning, or an aspect thereof, is to be identified with one or another psychological role associated with an expression. It has been proposed that the meaning of a sentence as used by a particular speaker is or involves one or more of: its inferential role (reflected by the speaker's dispositions to make inferences from and to the sentence), its evidential role (reflected in what observations and experiences incline the speaker to accept or reject the sentence), and its probabilistic role (the function that sends a sentence S and a collection C of sentences to the subjective probability the speaker would assign S if he held all of C true). (Developments of such views are in Boer and Lycan 1986, Field 1977, Sellars 1954.) None of these things can sensibly be evaluated for truth or falsity. Those who champion such psychological accounts of meaning often hold that meaning is a two-factor affair, the other factor being truth-conditional. Typically, though not invariably, the two factors are held to be independent.

In part, the appeal of adding psychological role to truth conditions in an account of meaning is that it seems to reflect a genuine tension in our pretheoretic conception of meaning. Consider Putnam's fantasy (in 1975) that there is a Twin Earth as much like Earth as possible, save that something other than H₂O, call it XYZ, plays the role that H₂O plays on Earth: XYZ has all the sensible properties of H₂O; it is XYZ, not H₂O, that fills the seas, that people drink and wash with, etc. Putnam holds, and many concur, that "water" means different things on Earth and on Twin Earth, for here it refers to H₂O, while there it refers to XYZ. But many think that in some very important sense the word has the same meaning in both places, for someone transported to Twin Earth who was innocent of chemistry, it is felt, would not mean anything different by "water" there than he means here. If there are two factors to the meaning of "water"—a truth-conditional one (which varies between Earth and Twin Earth) and a psychological one (which is constant), both intuitions are partially vindicated.

A different kind of nonpropositional meaning is what is sometimes called "expressive" meaning. The idea of such meaning has its roots in the work of emotivists like A. J. Ayer and Charles Stevenson. According to Ayer, the role of ethical discourse is completely noncognitive. Utterances of sentences such as "Stealing is wrong" and "Friendship is good" are not assertions and do not express beliefs. Rather, they are expressions of attitudes of approval or disapproval. Uttering "Stealing is wrong" is doing the sort of thing one does when one shouts "Down with stealing!" or accompanies utterance of the word "stealing" with a disapproving shake of the head. Stevenson's somewhat more sophisticated take on such sentences is that uttering them both expresses a distinctive sort of approval and exhorts (or at least attempts to bring) the audience to share this approval.

Sentences whose role is clearly exhausted by the expression of attitude—"Boo!" "Liver—yuck!" "Damn!"—are not candidates for combining with connectives and quantifiers to form larger sentences. "If liver—yuck, then I won't make dinner" does not have a meaning, for it is not even a sentence. But sentences such as "Stealing is bad" quite obviously do combine with connectives and other sentences, and the results certainly do seem to be meaningful. It seems incumbent on any account of semantics to explain what the meaning of a sentence such as "Stealing is bad only if it causes pain to someone."

Geach (1965), expanding on points in Frege (1918/1952), objects that the emotivist cannot make any sense of the use of normative vocabulary in complex sentences, of embedded uses, as is sometimes said. Someone who utters "If failing Mary will make her sad, you shouldn't do it" need not be expressing disapproval of anything. Even if there is a way around this—one might invoke some sort of "conditional disapproval"—emotivist views make the fact that we give normative arguments an utter mystery. The argument "Borrowing and not returning something is bad; if that is bad, so is stealing; so stealing is bad" is valid—its conclusion follows from its premises. But it seems to be nonsense to think that a feeling of disapproval for stealing follows from a feeling of disapproval for borrowing and not returning and whatever attitude might be associated with the conditional above. "Following from," after all, is a relation normally defined in terms of preserving truth. But if this makes no sense, the idea that the argument is valid makes no sense in emotivist terms.

These considerations, incidentally, bear on the view of Austin mentioned above. The argument "If I promise to meet you, I will meet you; I promise to meet you; so I will meet you" seems obviously valid. But there is a sort of ambiguity, on Austin's view, in "I promise to meet you." Embedded in the antecedent of a conditional, it presumably does nothing but express the statement that its user promises to meet the addressee. Unembedded, it apparently does not do this, as one, in uttering the sentence, does not assert that one promises, on Austin's view; one simply promises. It thus seems like the sense of "I promise to meet you" varies across the two premises of the argument, and thus the argument is not valid.

Expressivists such as Simon Blackburn and Alan Gibbard have recently tried to respond to this sort of objection, giving accounts that (more or less) agree with the emotivist line about simple sentences like "Hooking up is good" and attempting to derive therefrom meanings for complex sentences in which normative vocabulary occurs. Blackburn (1993) agrees with the emotivist that sentences like "Stealing is bad" express motivational states such as attitudes of disapproval. But he aspires to give an account of the meanings of the full range of uses of normative vocabulary, including such sentences as "Mary believes that stealing is bad" and "It's true that stealing is bad." The account is to be one that systematically assigns, to complex sentences, complex attitudes—typically in one or another way compounded out of the attitudes expressed by simple sentences. The sentence "If borrowing and not returning something is bad, then so is stealing it," for example, expresses a commitment to either tolerating borrowing and not returning, or to disapproving stealing. Such a view would allow us to characterize validity in terms of preservation of commitment—an argument is valid just in case it is impossible to fulfill the commitments associated with premises without fulfilling those associated with the conclusion.

Gibbard (1992, a recast of 1990) suggested that normative sentences—not just sentences from morality, but sentences about what is or is not rational—absorb their meanings holistically from their relations to "immediate motivations," that is, to the states one expresses if one thinks to oneself "Do/Don't do that now!" The idea, roughly put, is that just as complex statements get their truth-conditional content from their inferential relations to sentences expressing observations, so normative statements absorb their content from inferential relations to sentences expressing immediate motivations. Gibbard suggests that the meaning of a normative sentence (including complex combinations of normative and nonnormative elements) can be represented as a set of "factual-normative" worlds, which are pairs of possible worlds and systems of norms. The idea, again roughly, is this. A simple factual statement holds at world w and norm n if it is true there. A simple normative statement such as "That is bad" (whose connection with "Don't do that!" is obvious) holds at w and n provided that n forbids the act referred to. With this as a basis, one can use standard techniques to assign sets of factual-normative worlds to compound sentences.

One might argue with Blackburn and Gibbard about the details of their approaches, worrying, for example, that Blackburn helps himself without justification to the idea that there is a distinctive sort of moral disapproval. Yet it would seem that something along the lines of Blackburn's or Gibbard's story must be correct. Here is why.

Forget about claims about morality, rationality, or other obviously normative concerns. Think instead about what is going on when we talk about talk that obviously aspires to be true or false—about what happens when one person says "Jo is bald" and another says "That's not true," or when someone says "The sentence on the board isn't true." It seems obvious that such talk can get it right without being true. If the sentence on the board is a liar sentence, one thing that we know about it is that it is not true. We can, after all, prove that it is not. But paradox ensues if we take this thing we know—that the sentence is not true—to be true. After all, if what we know—that the sentence is not true—is true, then, since the sentence says just that—that it is not true—what the sentence says is true. So what we know is false. But one cannot know something that is false. Similarly, if vague predicates are neither true nor false of their borderline cases (and surely this is the most plausible thing to say about them) and Jo is borderline-bald, then while the person who utters "Jo is bald" says something, what he says is not true. But if it is true that the sentence is not true, then (since what is not true is false), "Jo is bald" must be false. But since Jo is borderline bald, "Jo is bald" cannot be false either.

What should we make of this? Well, for one thing, when we say, referring to the liar, "That is not true," we should not be understood as asserting something, that is, committing to its truth. Rather, we are performing the sui generis speech act of denial, where (roughly put) denying a potential truth bearer is the appropriate thing to do if it is not true ("not" being used here to deny). This sort of thing applies quite generally to uses of other logical connectives. Sometimes, for example, when someone utters "A if and only if (iff) B," they are to be understood as asserting the material equivalence of A and B. But when we say things with the form " 'S ' is true iff S " and S happens to be a liar sentence, we are not to be understood as asserting anything. Rather, we are performing an act that is apt if the claims connected by "iff" have the same (perhaps non-truth-conditional) status.

When we utter sentences, we perform different sorts of speech acts. Sometimes we assert, sometimes we deny, sometimes we perform the sort of act just mentioned. And when we perform such acts, we incur various commitments. For example, assertion commits us to the truth of what is asserted; denial of a potential truth bearer commits us to the nontruth thereof. Sentence-compounding devices, at least on some occasions, contribute not to sense, by (for example) expressing truth-functional negation, but to force. In the case of "not," for example, one sometimes signals that one is denying, where to deny S is to commit to the inaptness of whatever commitment is associated with uttering S.

Think of the simplest sentences of one's language as vehicles for performing speech acts, each such act involving its own distinctive kind of commitment, each commitment having its own conditions of appropriateness and inappropriateness. Annexing words like "not" and "if" to sentences yields (when the connectives signal force) sentences that are vehicles for performing speech acts with their own distinctive kinds of commitments, their own aptness conditions. Compounding sentences with several connectives playing the role of force indicators produces a sentence that can serve as a vehicle for performing a complex speech act determined by the meanings of the constituent sentences and the force-indicating meanings of the connectives. Uttering "If S is a liar sentence, then it is not true," for example, performs an apt speech act if it is apt either to deny that S is a liar sentence or to deny that S is true.

Beyond an account of sense or reference, a theory of meaning for a language—at least one component of such a theory—must tell the story of how the acts and commitments associated with the parts of a complex sentence determine the act for which the complex sentence is a vehicle, the commitments one incurs with the act, and the aptness conditions of such commitments. (For the beginning of such a story, see Richard 2006.) Such a story generalizes the sort of ideas Blackburn had. With such a story, one can see that logical validity, in its most basic sense, is preservation of commitment: An argument is valid provided that whenever the commitments associated with the premises are apt, so are those associated with the conclusion.

It was mentioned above that there was something importantly right about Gibbard's and Blackburn's accounts of normative discourse. What is important and surely right is not their view of the nature of the acts performed and commitments incurred in normative utterances. Perhaps those accounts are on the right track, because normative discourse is expressive, not truth-apt. Perhaps they are wrong, and normative discourse is no less truth-evaluable than a stock-price quotation. What is important is the insight that validity (and the other properties we associate with rational discourse) are not the exclusive property of truth-conditional discourse. Sometimes meaning and validity are to be explained in terms of truth conditions. But this is not the only case—it is but a special case.

See also Meaning.

Bibliography

Austin, J. L. How to Do Things with Words. 2nd ed. Cambridge, MA: Harvard University Press, 1962.

Ayer, A. J. Language, Truth, and Logic. London: Victor Gollancz, 1936.

Bach, Kent. "The Myth of Conventional Implicature." Linguistics and Philosophy 22 (4) (1999): 367–421.

Blackburn, Simon. Essays in Quasi-Realism. New York: Oxford University Press, 1993.

Boer, Stephen, and William Lycan. Knowing Who. Cambridge, MA: MIT Press, 1986.

Field, Hartry. "Logic, Meaning, and Conceptual Role." Journal of Philosophy 74 (1977): 379–409.

Frege, Gottlob. "Negation" (1918). In Translations from the Philosophical Writings of Gottlob Frege, edited by Peter Geach and Max Black. Oxford, U.K.: Blackwell, 1952.

Geach, Peter. "Assertion." Philosophical Review 74 (1965): 449–465.

Gibbard, Alan. "Reply to Blackburn, Carson, Hill, and Railton." Philosophy and Phenomenological Research 52 (4) (1992): 969–980.

Gibbard, Alan. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press, 1990.

Grice, H. P. "Logic and Conversation" (1967). In his Studies in the Ways of Words. Cambridge, MA: Harvard University Press, 1989.

Karttunen, Lauri, and Stanley Peters. "Conventional Implicature." In Presupposition, edited by Choon-Kyu Oh and David Dinneen. New York: Academic Press, 1979.

Potts, Christopher. The Logic of Conventional Implicatures. Oxford, U.K.: Oxford University Press, 2005.

Putnam, Hilary. "The Meaning of 'Meaning.' " In his Mind, Language, and Reality. New York: Cambridge University Press, 1975.

Richard, Mark. Beside Truth. Oxford: Oxford University Press, 2006.

Sellars, Wilfred. "Some Remarks on Language Games." Philosophy of Science 21 (1954): 204–228.

Stevenson, Charles. Ethics and Language. New Haven, CT: Yale University Press, 1944.

Mark Richard (2005)