2
Literal vs. non-literal meaning
In Chapter 1, we said that the difference between semantics and pragmatics was essentially the difference between convention and intention: Semantic meaning is conventional meaning, the meaning that a word takes with it wherever it occurs and whoever is using it. The word chair has a conventional meaning that speakers of English have more or less agreed on implicitly, even if we might disagree about the details—for example, whether a three-legged stool counts as a chair. And from this conventional meaning arise a range of metaphorical meanings, such as the chair of a committee or an endowed chair in a university. But these too are conventional, and appear in dictionaries. When we use one of these conventional meanings in a particular context, however, our intention adds another range of meaning: If I ask you to bring me a chair, whether or not a stool is appropriate might depend on whether I need to kill a spider on the ceiling (and need a chair to reach it), or whether I’m going to sit down to play the guitar (in which case a stool might be just the right thing), or whether I’ve brought in an elderly guest and want her to feel comfortable (in which case a stool is not the right thing at all). This intentional meaning is pragmatic. But the line between convention and intention is not at all clear-cut. In this chapter we will look at several ways of distinguishing between types of meaning, all with an eye to narrowing down what we mean when we talk about meaning—and what it is we’re talking about when we talk about pragmatics.1
Natural and non-natural meaning
Philosopher H. P. Grice established an important distinction between two types of ‘meaning’ (Grice 1957). Consider the following examples:
(1) 
(2) 
In each of the examples in (1), the ‘meaning’ in question is natural, unintended, and nonconventional. By ‘nonconventional’ I mean that there is no agreed-upon, prearranged convention by means of which society has decided that, for example, high temperatures will indicate illness; it’s just a fact about the way our bodies work. There is no convention by which dark clouds inform us of coming storms, termites inform us of trouble, smoke indicates the presence of fire, or a loud noise in the exhaust pipe indicates a bad muffler. It is simply natural for clouds to indicate storms, for smoke to indicate the presence of fire, and so on.
In (2), on the other hand, convention is involved. By ‘convention’, I mean (there’s that word again!) a relatively fixed, albeit tacit, societal agreement—a practice or viewpoint that members of a society share simply by virtue of being members of that society. So for English speakers, a convention exists by which the word brusque means ‘abrupt’; and while you might certainly choose not to follow this convention, it limits your ability to participate in this society. Recall the well-known encounter between Alice and Humpty Dumpty:
(3) 
The humor in this passage, of course, comes from how very wrong Humpty Dumpty is. The semantic meaning of a word has its basis in convention rather than in individual intention. Words only work for purposes of communication because we agree on their basic meanings; we’re not free to assign them any meaning we wish. The only intention here is the default intention to share in a society which has adopted English as its primary means of communication. The remaining examples in (2) are similar: Our society has adopted conventions by which a red octagon means ‘stop’ and a green light on a credit card reader means that the machine is ready to accept the credit card; German society conventionally uses the word Kopf to mean what English speakers mean in using the word head; and in America, the thumb-to-forefinger gesture conventionally means ‘okay’. The examples in (1) illustrate what Grice called natural meaning, whereas those in (2) illustrate what he called non-natural meaning.
So far, so good. But then things get murky. Consider these examples:
(4) 
One could argue that (4a) is actually ambiguous, depending on whether the smile is an automatic, uncontrolled result of the happiness resulting from winning or an intentional smile meant to convey, essentially, ‘Hey, I won!’ In (4b), there’s no automatic, natural connection between high temperatures and calls to doctors; but neither is it quite a convention. If pressed, I’d have to say that the high temperature naturally means the person is sick, as in (1a), and sickness in turn naturally means an intervention is needed; and that need for intervention conventionally (hence non-naturally) means that a call to the doctor is in order. Example (4c) is similarly a multistep relation; it’s conventional for flashing red lights on certain devices to mean the power has gone out, and once that convention is in place, the fact that the light is flashing red at this moment means naturally (automatically) that the power has gone out. In (4d) and (4e) we have cases of ‘meaning’ that are directly intentional on the part of the speaker. In (4d), the speaker is setting up a nonce convention between speaker and hearer. In (4e) the word mean could in fact be replaced by the word intend.
Conventional and intentional meaning
So we’ve seen that ‘meaning’ can be an automatic relationship with neither convention nor intention behind it (as with clouds mean rain), or it can be a conventional relationship in which some group has agreed that one thing will stand for another (as with the word brusque means ‘abrupt’), or it can be simply the intention of an individual in a context (as with I didn’t mean to hurt your feelings), or it can be some combination of these. We have made a first cut between natural meanings of the first sort and the non-natural meanings involved in the rest of the cases. Within these non-natural meanings, however, there is obviously at least one more distinction to be made: We need to distinguish between conventional and intentional meanings.
Now, at one level you could object that conventional meaning is also intentional: Like Humpty Dumpty, I could perfectly well decide that I’m going to use the word glory from now on to mean ‘a nice knock-down argument’. So at some level, by using the word glory to mean the same thing everybody else uses it to mean, I’m intentionally participating in the norms of our society. But at another level, it isn’t my own intention that has set up the relationship between the word glory and its meaning (roughly, ‘magnificence’). So there’s a distinction to be made between a speaker’s intention to use a word in the conventional way, and the historical processes by which it has come to have its conventional meaning, which usually aren’t the result of any one person’s intentions. In this latter sense, then, my using glory to mean ‘magnificence’ isn’t a matter of intention; it’s a matter of convention.
At this point we approach the distinction between semantics and pragmatics. As a first pass, we could say that semantic meaning is conventional, while pragmatic meaning is intentional. Most of our utterances have both a conventional and an intentional component. Recall the testimony of James Comey before the Senate Intelligence Committee in June of 2017, concerning a conversation between him and President Donald Trump, when he was Director of the FBI. As reported in the press:
(5) 
Here we can see clearly the distinction between what has been conventionally stated and what Comey believes to have been intended. Note, of course, that others saw the intention differently:
(6) 
The conventional, societally shared meaning of the phrase I hope X is roughly that the speaker would prefer X to some alternative. And this is the meaning that Trump Jr. is relying on in saying “I hear ‘I hope nothing happens.’” The second half of his interpretation, ‘but you have to do your job,’ goes beyond this to the assumed intention of the speaker: Trump Jr. is claiming that this statement from the President of the US, in this context, would naturally include the intention that Comey do his duty; none of that is stated conventionally, but Trump Jr. is arguing that it’s contextually evident. Comey explicitly argues for a quite different pragmatic meaning: “I took it as a direction… I mean, this is a president of the United States with me alone saying, ‘I hope this.’” Comey is saying that this statement from the President of the US, in this context, would naturally include the intention of directing Comey to do what the President has said he hopes he’ll do: let Flynn go. Comey is essentially arguing that when the President of the United States privately tells a subordinate I hope you can do X, the clearly intended message is ‘do X’.
Major news events don’t always hand us such a clear analysis of distinct pragmatic interpretations of a single semantic meaning. Note also that what makes the difference in this case is a specific aspect of the context: the speaker of the utterance in question. Comey argues that because the speaker is the President, who is a particularly powerful individual (not to mention Comey’s boss), when he says I hope you can do X it carries the force of a directive to do X. Trump Jr., on the other hand, emphasizes the integrity of the office in his interpretation: Because the speaker is the President, he reasons, he would not issue a directive to do anything unethical or illegal, hence the interpretation ‘you have to do your job’. All of which is to say that a chief difference between what’s conventional and what’s intentional is the role of context.
Context-dependent and context-independent meaning
The context for an utterance includes everything about the situation in which the utterance takes place: not just where and when it happens, but also the identities of the speaker and the hearer(s), their shared background and beliefs about each other (and about each other’s beliefs, etc., ad infinitum), the assumed goals of the interaction, and so on. Because of the importance of context in helping a hearer to interpret a speaker’s meaning, we make a distinction between a sentence and an utterance. A sentence is an abstract linguistic object that expresses one or more propositions, and it may or may not ever be uttered. An utterance is the use of a linguistic expression (word, phrase, sentence, etc.) in a context. The rules of the English language make it possible to construct a literally infinite number of distinct sentences, which means that not all of them can ever actually be uttered. With a moment’s thought, you can easily construct a sentence that has never been uttered before. Here are a few sentences that have probably never before in the history of the universe been uttered:
(7) 
I can be fairly sure these haven’t been uttered before because they’re a bit silly (and in any case, they’ve NOW been uttered, since one way of ‘uttering’ a sentence is to write it), but if you pay careful attention to the sentences you encounter in a given day, you’ll realize that a fairly large proportion of them have probably never been uttered before (like, for example, this one). The crucial thing about language is, of course, the fact that you can understand never-before-uttered sentences like these—and the reason you can do so is that you share the conventional rules and vocabulary of English and know how these rules and vocabulary are used for putting together meaningful sentences.
There are, then, an infinite number of sentences of English; some of them have been uttered, some haven’t, and some have never even been thought of and never will be. A sentence that has never been uttered still has meaning by virtue of the conventions of grammar and vocabulary that English speakers in general share. This meaning is generally context-independent; it does not depend on the context in which a sentence is uttered (or indeed whether it’s ever uttered at all), except to the extent that context helps us pick out which of a set of conventional meanings was intended (e.g., whether a speaker saying Phyllis bought new glasses means eyeglasses or drinking glasses).
So once a sentence is uttered, the context matters. In (5), part of the context of the utterance I hope you can let this go is the fact that the US President has uttered it, and that context affects Comey’s interpretation. If I tell a casual acquaintance I hope you can make it to my party, the words I hope have that same context-independent semantic meaning but a very different context-dependent pragmatic force; in the absence of some other reason to assume coercion, the statement would be taken as an invitation, not as a directive.
In short, one easy way to slice up the meaning pie would be to say that the difference between semantics and pragmatics is precisely the difference between context-independent meaning and context-dependent meaning. This is very close to saying it’s the difference between convention and intention, but it’s not quite the same. It’s fair to equate context-independent meaning with conventional meaning; both boil down to what a word or sentence conventionally means regardless of context. But intention is a bit more slippery, because we can never know a person’s true intention. To go back to the example in (5) and (6), we see that we have a single utterance in a single context, but two different interpretations of the speaker’s intention. What was Trump’s actual intention? Only Trump knows for sure. A given hearer will make their best effort to determine what the speaker intended, and their efforts will be guided by a combination of conventional meaning and contextual factors (along with some general principles we’ll introduce in the next chapter), but ultimately they will arrive only at an inferred meaning, which may or may not be the intended meaning. And simply asking the speaker is no help; we can ask President Trump what he intended when he said I hope you can let this go, but it’s clear that some people will believe his answer and others will not.
So there’s a distinction between what is conventional and what is intended, and there’s also a distinction between context-independent meaning and context-dependent meaning. Conventional meaning corresponds to context-independent meaning; even a word like I has a context-independent conventional meaning (something like ‘the person speaking’) that, in turn, combines with the context to determine the intentional meaning (cf. Kaplan 1989). But intentional meaning does not correspond straightforwardly to context-dependent meaning, because intentional meaning is speaker-based (it’s the speaker who intends the utterance to mean a particular thing) whereas context-dependent meaning is essentially two-pronged: There’s the meaning that a speaker intends the hearer to arrive at after considering the context, and then there’s the meaning that the hearer actually does arrive at, which can be quite different, and there’s no way to bridge that divide. Pragmatics as a field is interested in both—the speaker’s intention and the hearer’s interpretation—but it is also deeply interested in the disconnect between the two. As you read this book, it will be crucial to keep in mind that intention and interpretation can differ widely, and indeed this is one of the chief sources of miscommunication.
Interestingly, the conventional meaning in (5) is something that both sides can agree on: Presumably we can all agree that President Trump’s words meant that he did in fact, strictly speaking, hope that Comey could ‘let this go’; that is, the conventional (i.e., ‘literal’) meaning of the sentence I hope you can let this go is: Trump hoped Comey could let it go. And presumably this semantic meaning is true (he did hope Comey could do that). There are at least two pragmatic (nonconventional, context-dependent, inferred) rivals for the intended meaning: One is ‘I direct you to let this go’, and the other is ‘I hope that you are able (e.g., not forbidden by the law) to let this go’. These two possibilities have a different relationship to truth; that is, while it is true in either case that Trump hopes Comey ‘can let this go’, it’s not quite so clear that we can make a true/false statement concerning whether Trump directed Comey to do so.2 To explore this a bit further, we need to talk about the relationship between meaning and truth, which turns out to be central to the distinction between semantics and pragmatics.
Truth-conditional and non-truth-conditional meaning
There is a long tradition in linguistics of equating the difference between semantics and pragmatics with the difference between those aspects of a statement’s meaning that affect its truth and those aspects of its meaning that do not. For example, to move to the other side of the political aisle, you might recall that when President Clinton was responding to allegations concerning his relationship with Monica Lewinsky, at one point he quite strongly stated (8):
(8) 
The question is, did he speak the truth? And of course—without going into the graphic details—it all depends on the meaning of ‘sexual relations’, and this is a point on which people disagree. (See Tiersma 2004 for discussion of the issue and the court’s interpretation of the term.) What that means is that the supposedly conventional meaning of ‘sexual relations’ is not 100% conventional, in the sense that it is not 100% shared. But you may be surprised to learn that few conventional meanings actually are. Recall the discussion of lexical semantics in Chapter 1: If you ask a group of people whether a stool, a love seat, and a bean bag chair each count as a chair you’ll find you get quite a range of responses—and if you ask them whether a hot dog is a sandwich, be prepared for a heated argument! The fact of the matter is that ‘conventional’ meaning is not as universally shared as the word ‘conventional’ might lead you to expect.
Nonetheless, for a given semantic meaning of chair, that semantic meaning will determine whether a sentence like (9) is true:
(9) 
And likewise, whether (8) is true or false depends on the semantic meaning of ‘sexual relations’—but this is a point on which reasonable people disagree. In the case of (8), a great deal hinged on this question, in particular whether the President had lied to the American people. And because the semantic question is one on which reasonable people differ, it’s a difficult question to answer.
Compare the utterance in (8) with another utterance on the same topic:
(10) 
This is a statement President Clinton made in testimony before a Grand Jury. The context is that his lawyer had previously stated, “Counsel is fully aware that Ms. Lewinsky has filed, has an affidavit which they are in possession of saying that there is absolutely no sex of any kind in any manner, shape or form, with President Clinton.” The questioner then asserts to Clinton that “That statement is a completely false statement,” and asks Clinton whether he agrees. Clinton responds:
(11) 
In short, he’s arguing that because the relationship had ended months earlier, the statement there is absolutely no sex of any kind…was true at the time it was uttered, because the word is is in the present tense. Semantically, this is accurate. Pragmatically, however, one could argue that it’s purposefully misleading: The information that is really wanted is whether there had been such a relationship, and Clinton’s lawyer is purposefully dodging this question, answering with respect to the literal truth despite knowing that it will mislead the hearer. Clinton, then, vouches for the truth of the statement on the grounds that part of the semantic meaning of the word is is the present tense, and the relationship was in the past.3
What this example shows us is the difference between those aspects of meaning that contribute to the truth of an utterance and those aspects of meaning that might not contribute to its truth but nonetheless contribute to the intended (or inferred) meaning. To slice through a lot of extraneous detail in (10)–(11), a person saying There is no relationship in the hopes that the hearer will also infer ‘there was no relationship’ has said something that is semantically true but pragmatically misleading.
As noted in Chapter 1, the conditions under which a proposition counts as true are its truth-conditions, and whether that proposition is true in a particular world is its truth-value in that world. Truth-values are basically binary; a proposition is either true or false (though we’ll see another possibility when we talk about presupposition in Chapter 7). Aspects of meaning that affect whether a given sentence is true are truth-conditional aspects of meaning. So one way of looking at the difference between semantics and pragmatics is to say that semantic meaning is truth-conditional and pragmatic meaning is not: I hope you can let this go is true if and only if the speaker actually hopes the hearer can let it go, and There is no relationship is true if and only if at the time of utterance no relationship exists. These are the semantic meanings of these utterances. Inferred meanings such as ‘I direct you to let this go’ and ‘There was no relationship in the past’ are different, and count as pragmatic meaning; if one of them turns out to be false, it doesn’t affect the literal truth of the sentence. One problem with this is that it’s vastly more straightforward to determine the truth-conditions of a statement than those of, say, a question or command (though, needless to say, various approaches have been advanced to extend a truth-conditional account to these other expression types).
There is a huge overlap between conventional meaning and truth-conditional meaning; in the vast majority of cases, the conventional meaning of a word or sentence corresponds to the meaning that affects truth-conditions. But not always. Consider (12):
(12) 
Let’s examine two words in the above sentence: abundant and yet. The word abundant means something like ‘copious’ or ‘plentiful’; the word yet, in its use as a conjunction, means something like ‘in contrast’ or ‘despite that’. These meanings are conventional, in that it’s hard to imagine a use of abundant in which it wouldn’t mean ‘copious’ or a use of yet in which it wouldn’t mean ‘despite that’.4 Nonetheless, the truth-conditional status of these meanings differs. If the author’s appreciation and gratitude were not, in fact, plentiful, then (12) is false. But if there’s no contrast between working against powerful forces and feeling grateful, most people would agree that (12) is still true; it just becomes odd to use the word ‘yet’. So compare (12) with (13):
(13) 
There’s certainly no contrast between being helped by loving and compassionate people and feeling appreciation and gratitude, but as long as the author was indeed helped by such people and did indeed feel such emotions, (13) is true; the odd use of yet doesn’t render it false. (Nonetheless, the reader typically would try to find a reading on which the contrast is present; perhaps there’s some reason why encountering love and compassion would lessen one’s gratitude?) Thus, the ‘copious’ meaning of abundant contributes to the truth-conditions of a sentence, but the ‘contrast’ meaning of yet does not. That’s what it means to say that these bits of meaning are truth-conditional and non-truth-conditional, respectively.
This leaves us with a slight disconnect between several possible ways of distinguishing between semantics and pragmatics: One could say that semantics is conventional and pragmatics is intentional, which is tidy and easy to remember and generally true, but has the downside that convention and intention aren’t quite opposite sides of the same coin, and that this leaves out the hearer’s interpretation. Or one could say that semantics is context-independent and pragmatics is context-dependent, which is a clunkier but more precise way of saying much the same thing; but in both cases we’ve lost the connection between semantic meaning and truth—which makes it hard to address questions of lying vs. truth-telling, perjury, and what has been ‘literally’ said. Or one could say that semantics is truth-conditional meaning and pragmatics is non-truth-conditional meaning, which preserves the connection between semantic meaning and truth but loses the connection between semantic meaning and convention, so that we end up having to uncomfortably say that despite the fact that ‘contrast’ is an inescapable part of the conventional meaning of the word yet, it’s not part of its semantics.
What to do? At first (and possibly second and third) glance, the truth-conditional view of semantics is at a distinct disadvantage, because there’s so much meaning associated with a typical utterance that doesn’t seem truth-conditional at all, as we saw in the examples given. But truth-conditional semantics grew out of a long history of work in philosophical logic, and philosophers in the mid-twentieth century weren’t eager to throw the baby out with the bathwater. Was there some way to preserve truth-conditional semantics while accounting for the difference between logical ‘truth’ and the richer meanings associated with real-world communication? That’s the problem that philosopher H. P. Grice took on. You may remember Grice from the distinction between natural and non-natural meaning we have discussed. He was one of the most influential figures in pragmatic theory, if not the most influential figure; and he is best known for his Cooperative Principle, which attempts to bridge the gap between semantic and pragmatic meaning, and which we’ll discuss in the next chapter.