The semantics of presupposition

↵ Back to class homepage

Way back in the "Truth-conditional semantics" module, we introduced an example "The current king of France is bald". We discussed how some people say this sentence is neither true nor false, because there is no current king of France and thus we can't judge whether that [non-existent] person is bald.

In pragmatic and semantic theory, we can say that this sentence presupposes that there is a current king of France. (Or, equivalently, the proposition "There is a current king of France" is a presupposition of the sentence "The current king of France is bald.") In other words, the main focus of the sentence is not to claim that there is a current king of France, but the sentence expresses that information anyway in the background, and the main claim of the sentence depends on this background claim in order to make sense (if this background claim is not true, the sentence as a whole won't make sense).

Here's another example. In an early scene in the novel A Desolation Called Peace by Arkady Martine, two of the main characters, Mahit and Yskandr, are having a meeting with a nefarious high-level government official. Yskandr believes that the official sabotaged Mahit and Yskandr's work recently; Mahit is not so sure, and has told Yksandr that. Anyway, they aren't originally sure why this official has called them to a meeting, but over the course of the conversation Mahit deduces the purpose of the meeting, and Mahit and Yskandr quietly have the following discussion:

Here, Mahit has not directly said that she believes the official sabotaged them. But by referring to "the sabotage" she has presupposed that the sabotage in question exists / did happen. It wouldn't make sense for Mahit to say "The official wants to know if the sabotage worked" if she didn't believe there was a sabotage at all.

For another example, consider a bittersweet scene from the video game Horizon: Zero Dawn (minor spoilers). Here, a character named GAIA is talking to a character named Elisabet about what Elisabet would have liked her children to be like if she had children.

Here, Elisabet's utterance actually conveys two big pieces of information. One, the proposition that she would want her child to be curious and willful and unstappable and with compassion to heal the world, is the response to GAIA's direct question. But another piece of information she conveys is that if Elisabet had a child, she would want to have a daughter. This is sort of background information in Elisabet's utterance; she communicates this information without making it the main focus of what it's saying, she just sneaks it into the background. In short: while Elisabet says that she would have wanted her child to be curious/willful/unstoppable/compassionate, she also presupposes that she would have wanted her child to be a girl.

Presupposition is one of the oldest, most discussed, and most complicated topics in pragmatics and semantics. There are many examples of kinds of sentences that presuppose some background information (see Levinson, chapter 4, for a huge list). For example, "What the world needs now is love, sweet love" presupposes that the world needs something—if the world didn't need anything, it wouldn't make sense to say that sweet love is what the world needs. And "Doctors told me that if I didn't give up these foods, I'd probably keep shitting my pants" presupposes that the speaker had been shitting her pants—if she wasn't shitting her pants already, then she couldn't "keep" shitting her pants. (It also presupposes that she was eating "those foods"; they couldn't tell her to "give up" something she already wasn't eating.)

What's a presupposition?

One of the challenges with discussing presuppositions is that there's no clear definition of what they are; different theoretical approaches have different explanations. Some people say a presupposition is something that has to be true for you to be able to judge whether the utterance is true. As we've seen above, we can't decide whether or not the current king of France is bald if there is no current king of France whatsoever. Likewise, we can't decide whether the thing that the world needs now is love, if we don't believe the world needs anything anyway. A useful way to notice presuppositions is to imagine how you would respond to a question that presupposes something false. For example, if someone asks you "Have you stopped shitting your pants yet?" when you already were not shitting your pants, you probably wouldn't respond "yes" or "no"; your response would be something more like, "Huh? That question doesn't make sense—I wasn't shitting my pants to begin with."

In this section we'll examine in a bit more detail the following three properties of presuppositions:

Insensitivity to negation

The most famous feature that sets presuppositions apart from truth-conditional semantic meaning (i.e., entailment) is that [most] presuppositions are not affected by negation or other environments that reverse entailments. This probably sounds a little bit abstract, but we can see it more clearly with some examples:

Long story short, presuppositions are things that some utterance seems to communicate, but that are not part of the literal meaning of the sentence. (But, as we will soon see, this definition is not sufficient, because it doesn't explain how presuppositions are special and different from implicatures; implicatures are also something that's not part of the literal meaning of the sentence.

This negation test can help us understand another presupposition example. In an episode of BoJack Horseman, BoJack's friend (and on-again off-again girlfriend) Princess Carolyn has just criticized him for not remembering anything about their relationship, and he replies, "That's not true. I remember the first time we met. I went to see Marv. You were there at the desk in front of his office. I thought you were kind of cute, and you said, 'Hey, it's good to see you again!' Huh, so I guess we'd met before that." Here, remembering that Princess Carolyn said "Hey, it's good to see you again!" makes him realize that actually was not the first time they'd met... but why? She didn't explicitly say they had met before that. Instead, she presupposed it: "it's good to see you again" presupposes that the speaker has met the addressee before. You can tell this is a presupposition and not an entailment, because it is not affected by negation: "it's not good to see you again" would still presuppose that the speaker has seen you before (no one would say "it's not good to see you again" to mean "it's good to see you and I've never seen you before"!).

The examples we saw at the beginning of this module also pass the negation test. "What she wants is to know if the sabotage worked" and "She doesn't want to know if the sabotage worked" both express that there was a sabotage, regardless of the negation. And "If I'd had a child, I would not have wanted her to be curious, willful, unstoppable, and compassionate" would still express that Elisabet wanted her child to be a girl, even with the negation. And "Doctors told me I won't keep shitting my pants" still means the speaker has been shitting her pants before.

Utterances with false presuppositions are weird rather than false

Another helpful way to identify presuppositions is that an utterance whose presupposition is false often doesn't make sense. A presupposition is some background information that needs to be true for the utterance to make any sense; if an utterance presupposes something that is false, the utterance itself supposedly can't even be true or false, it's just a weird or nonsensical utterance. We can see this with all of our examples so far. "The current king of France is bald" seems like a weird sentence because there is no current king of France. "That's what she wants, to know if the sabotage worked" wouldn't make sense if there was no sabotage at all. "If I had a child, I would have wanted her to be curious, willful, unstoppable, and compassionate" wouldn't sound right if the speaker didn't want their child to be a girl. "Doctors told me if I don't stop eating these foods I'll probably keep shitting my pants" would sound like a crazy thing to say to someone who hasn't been shitting their pants to begin with. And "It's good to see you again!" is not a normal thing to say to someone you've never seen before. In each of these cases, the reason the sentence sounds weird is that it presupposes something which is false.

One way to notice the difference between "weird" and "false" is to imagine how you would respond to an utterance that has a false presupposition. I bet you often would not reply to these utterances by saying "no" (if someone you never met said "It's good to see you again", would you say "No, it isn't"?); often the most natural way to reply is to just point out that the utterance is weird and the presupposition is wrong (e.g., I might respond, "Huh? We've never met before!").

Note that this is different from entailments; when an utterance has an entailment that is false, the whole utterance is just plain false. For instance, if someone utters something that has a false entailment (such as saying "That's a black cat" [which entails that that's a cat] while pointing at a dog), it would be pretty normal for you to just say no (e.g., "No, it's not").

Backgroundedness

Another major property of presuppositions is that they tend to be expressed as background information, rather than the main focus of the utterance. Pragmaticists call the main focus of the utterance the "at-issue content" (i.e., the information that is "at issue", or the information that is really being discussed), and say presuppositions are "not-at-issue". In other words, when someone says something with a presupposition, they are often assuming (or at least acting as if) the presupposition is an uncontroversial thing that everyone in the conversation will accept, whereas the "at-issue content" is new information they are contributing to the conversation. We can see this most clearly in the example from the discussion between GAIA and Elisabet. The content that's "at issue" is the response to GAIA's question, so Elisabet says that as an entailment; the extra presupposition about the gender of her child is just extra information that doesn't have a bearing on the question at issue.

While this "backgroundedness" thing is technically a feature of presuppositions, it's hard to use it as a reliable diagnostic for presuppositions. The reason it's hard to use as a diagnostic is that speakers often manipulate it. In other words, speakers often strategically use presuppositions to try to sneak new information into the conversation, and get it accepted by others, without making it the focus of attention. (Think of, e.g., someone who says something like "I'm having lunch with my boyfriend today" in order to brag that she has a boyfriend—here, even though the fact that she has a boyfriend is a presupposition, it might still be the "main" thing that she wants to express, and the rest of the sentence might just be an excuse for her to sneak that presupposition in.)

This property of presuppositions is really more of a pragmatic than a semantic thing, and we will have more to say about it in the next module.

Presuppositions vs. conventional implicatures

If you've been thinking about implicatures while you've been reading this, you might have noticed that presuppositions sound very similar to conventional implicatures! Conventional implicatures, like presuppositions, communicate some extra background information that's not part of the literal semantic meaning. When a conventional implicature is false, it makes an utterance seem "weird" rather than false. Furthermore, conventional implicatures also do not change under negation! For example, as we've seen in the "Types of implicatures" module, an utterance like "These cookies are English, but they're good" communicates that English stuff is usually not good, and this comes from a conventional implicature rather than from the literal semantic meaning. But saying "It is not the case that these cookies are English but good" seems to have the same conventional implicature; this utterance seems to be saying that these cookies are not English-but-good (either because they're not English, or they're not good), rather than saying "These cookies are English and good and there is no contrast between those two properties." (It only seems possible to get that interpretation if you pronounce it with special contrastive stress, as in "These cookies aren't English but good... they're English and good!" Laurence Horn calls this sort of thing "metalinguistic negation", and it seems to work in a different way than normal negation—basically in this example it seems like the phrases "English but good" and "English and good" are acting like they're in quotation marks.)

For a similar example, we can consider pronouns that encode social heirarchy. For example, in some varieties of Mandarin, there are two second-person pronouns: 你 () for normal usage, and the polite 您 (nín), used to refer to someone who is socially "higher" than the speaker. (This is an example of the T-V distinction, where languages have different "polite" and "familiar" forms of second-person pronouns; this name comes from the fact that in Romance languages the "familiar" pronoun usually starts with T [e.g., tu in French] and the "polite" pronoun with V [e.g., vous in French].) So, for example, if you say 我明天把稿子交给您 ("I'll submit the draft to youpolite tomorrow"), you are communicating that the person you are speaking to you is socially "higher" than you (perhaps your boss or something). This will not change under negation; e.g., if you say 我不会明天把稿子交给您 ("I won't submit the draft to youpolite tomorrow"), you are still communicating that the person is socially higher (this utterance cannot be interpreted as meaning something like "I will submit the draft to you tomorrow and you are not socially higher than me"!).

The pronoun example above also illustrates the similarity between conventional implicatures and presuppositions. In fact, some people have analyzed the social connotation attached to the polite pronoun as a presupposition, and some have analyzed it as a conventional implicature (see Levinson, chapters 2 and 4). Both are non-literal meanings that are not reversed under negation.

One of the key differences that some researchers have claimed, though is that presuppositions are supposedly defeasible (cancellable). Recall that, as we learned in the the module on implicature diagnostics, conventional implicatures do not seem to be cancellable; trying to cancel them yields a very weird utterance ("These cookies are British, but they're good—and I'm not trying to suggest that British stuff isn't usually good"). On the other hand, presuppositions can be cancelled pretty easily and without very awkward-sounding utterances:

These examples are problematic, though, because they all involve presuppositions in negative sentences. Thus, it might just be another example of "metalinguistic negation". In that sense, these are maybe not so different from conventional implicatures, which also can be cancelled in metalinguistic negation, as we've seen above (e.g., "These cookies aren't English but good, because there's no contrast between being English and being good"). So maybe these are not a good example of the cancellability test. It's much harder to cancel a presupposition without metalinguistic negation; e.g., "You're going to keep shitting your pants, but you're not shitting your pants now" sounds like a contradiction.

Better evidence comes from the fact that presuppositions can be cancelled implicitly, without an explicit follow-up "because" sentence like in the examples above. Instead, presuppositions can just be cancelled by context and world knowledge—and this can be done without any metalinguistic negation. For example, before X often presupposes that X happened, but sometimes the rest of the sentence rules it out. "Before the bomb exploded, the building was successfully evacuated, so nobody got hurt" presupposes that the bomb exploded. But "Before the bomb exploded, Batman defused it" does not presuppose that the bomb exploded (because our world knowledge tells us that, once the bomb is defused, the bomb won't explode), and it also doesn't sound like a contradiction. This seems noticeably different from conventional implicatures, which can't be cancelled even by world knowledge (e.g., as we've seen before, "This is Chongqing style hotpot but it's spicy" just sounds like a weird sentence, because it implies that Chongqing hotpot is usually not spicy, which is false; crucially, our world knowledge that Chongqing food is spicy doesn't cancel that implicature, it just conflicts with it and makes the sentence sound wrong to us).

On this view, the social information communicated with polite vs. familiar pronouns seems more like a conventional implicature than a presupposition. If you use an inappropriate pronoun (e.g., using a familiar/informal pronoun to address someone who expects you to use the polite/formal pronoun, or vice versa), the context doesn't cancel the implicature that comes with the pronoun; rather, the implicature is still there, and that makes the pronoun noticeably inappropriate for the context. In fact, speakers know that, and take advantage of it to put distance between themselves and their listener, or to show their respect or disrespect to the listener, etc. For instance, my high school French teacher told me she once went to a restaurant and the waiter gave very poor service, so she addressed him with "tu" (the familiar pronoun) instead of "vous" (the formal pronoun, which is what would be expected in this situation) in order to show him how pissed off she was—and the waiter was shocked when he heard her say that. Crucially, in this situation, the real-world context (knowledge that she and the waiter are strangers and are expected to address each other with the formal pronoun) didn't cancel the familiar/informal connotation of "tu"; rather, that connotation was still there, which explains why the waiter was so shocked.

Note, however, that there is still lots of debate over what the differences between implicatures and presuppositions are, or whether there even is a difference. There are some theories that claim there's actually no such thing as presupposition, and that presuppositions are just special kinds of entailments or implicatures (see Zufferey et al., chapter 5, for a recent review; or Levinson, chapter 4.4, for a detailed account which argues that presuppositions are actually just a special interaction between entailments and conversational implicatures).

Video summary

In-class activities

Immunity to negation is not a perfect test for presupposition; some presuppositions are affected by negation. For example, consider anymore. A sentence like "Nobody watches Squid Game anymore" presupposes that people used to watch Squid Game. In many varieties of English, anymore is a "negative polarity item" and thus can only be seen in negative sentences like this (or in other downward-entailing environments like questions or the antecedents of conditionals). But in my home dialect, people can also use anymore in positive sentences, such as "A lot of people are watching Squid Game anymore" (this means something like "there didn't used to be a lot of people watching Squid Game, but now there are"). This utterance presupposes that people didn't used to watch Squid Game much; in other words, the presuppositions in the negative and positive sentences are the opposite.

Another, less exotic, example is questions. Levinson (chapter 4.3) claims that alternative questions (like "Is Newcastle in England or is it in Australia?") and wh-questions (like "Who watched the football game yesterday?") both have presuppositions that are sensitive to negation.

Have students discuss to figure out what the presuppositions of these questions are, and what happens to them under negation.


If it's hard for students to get started, you could get them started by first telling them what the presuppositions of these questions are supposed to be, and then letting them try out the negation test with them.

Specifically, Levinson claims that alternative questions presuppose that one or both of the alternatives are true—i.e., "Is Newcastle in England or is it in Australia? presupposes that Newcastle is in England or it is in Australia—and wh-questions presuppose that that the thing you're asking about exists—i.e., "Who watched the football game yesterday?" presupposes that someone watched the football game yesterday.

The latter case is easier to see being sensitive to negation: for example, if I'm in a classroom and ask "Who didn't watch the football game yesterday?", maybe everyone will raise their hands, and this wouldn't necessarily be surprising, because I don't think my question would have presupposed that someone watched the football game [at least, not someone in the context of that classroom]. Note that this is a different use of negation than the sort of ironic use, where e.g. someone asks who watched the football game and I reply "Who didn't watch the football game?", with stress, to suggest that almost everyone watched it.

For alternative questions, I guess we could negate "Is Newcastle in England or Australia?" (notice that this is a slight rewording of the original example) by saying "Is Newcastle not in England or Australia?" and this would not presuppose that it's in {England or Australia}. But notice that this example is pretty tricky because this negative question has two quite different interpretations (sort of like the negative wh-question above also did). One interpretation could be one of surprise: i.e., maybe I thought England was in Newcastle or Australia, but then I discovered some evidence that I was wrong, so I might be like, "Oh, wow! Is Newcastle actually not in England or Australia after all?" On the other hand, a different interpretation of this utterance could be sort of reinforcing something that should be obvious, or confirming something, or expressing incredulity. In English it's often more natural to do this with a contraction ("Isn't Newcastle in England or Australia?"), and it has pretty much the same meaning as putting "right?" on the end of a statement (i.e., "Newcastle is in England or Australia, right?").

Levinson (p. 179) offers the following example of an utterance with a lot of prepositions:

Let students try to find as many presuppositions in this sentence as they can.

If it helps, you can provide them a list of common presupposition triggers. The following is a non-comprehensive list:

As for an "answer" to this discussion topic, here are the presuppositions Levinson lists for the original sentence:

But note that this classification is not uncontroversial. For example, some sources treat appositives as conventional implicatures. If we go by the cancellability test, they certainly do seem more like conventional implicatures (or even entailments) than like presuppositions. If we take that view, then the fact that John is a good friend of the speaker might be a conventional implicature rather than a presupposition.

In the examples of implicatures we've seen throughout this semester, the implicature is usually figured out based on what the speaker said (e.g., for a typical quantity implicature, maybe a speaker says "Josh is smart" and a hearer thinks, "Hey, why didn't she say Josh is brilliant, she must believe he's not brilliant!").

But implicatures can also come from things a speaker presupposes without explicitly saying.

For example, in a scene in the China Miéville novel Un Lun Dun, the main character, Deeba, is talking through a translator to a guy named Mr. Claviger, who has some special headgear that she needs and that he won't give her. He's sitting up in a high place where she can't reach him, so she wants to make him angry enough to come down and attack her so she can steal his headgear. So she tells her translator to say, "Well, we don't want Mr. Claviger's headgear. Is he stupid? What sort of idiots does he think we are? Maybe WE aren't the idiots." This utterance outrages Mr. Claviger.

Clearly Deeba implicates here that Mr. Claviger might be an idiot. But how? "Maybe we aren't the idiots" does not actually say Mr. Claviger is an idiot. But it presupposes (through the use of a definite noun phrase with "the") that someone is an idiot. If Deeba and her friends are [maybe] not the idiots (this is entailed by what Deeba said), and there is an idiot around (this is presupposed by what Deeba said), then a relevance implicature will let us infer that Mr. Claviger might be the idiot.

Let's look at another, slightly more complicated, example, which involves irony. Irony, or sarcasm, is when someone not just says something they know is untrue, but they in fact mean to show that they think that thing is stupid or wrong. When a person uses irony or sarcasm, they're sort of "performing" the role of someone else saying that thing, and showing that they think anyone who would say that is stupid. For example, if I think pragmatics is really boring, I might say "Oh, pragmatics is sooooo interesting" to show that I think it's boring (and that I think the claim that pragmatics is interesting would be a really stupid thing to say). That's irony. Now on to our example.

The following short dialogue occurs between two law enforcement agents in an episode of Breaking Bad. Let's look at the dialogue first and then break down what it means.

To understand this example we need a bit of context about the story happening within the show, and a bit of context about US history. First the context within the show. These law enforcement agents have been trying to find and capture a notorious druglord who goes by the nickname "Heisenberg". Earlier in the show they arrested a guy, James Kilkelly, who admitted to being the druglord Heisenberg. But actually he wasn't; in fact, the real Heisenberg paid this guy to pretend to be him and go to jail in his place. In fact, Kilkelly had been doing that for years: basically his whole career is getting paid by rich guys so he can admit to crimes they actually did and then he can go to jail in their place. Hank, in the dialogue above, knows this guy's reputation, so he doesn't believe this guy is the real Heisenberg.

As for the US history part: Jimmy Hoffa was a famous Chicago labor union leader who disappeared in the 1970s. It is assumed that he was murdered, but his body was never found and police never figured out who killed him. In the decades since then, many wannabe tough guys who want fame and attention have claimed to be Hoffa's killer. (For example, the popular—but astoundingly boring—2019 film The Irishman is about one of those stories.)

With that background in place, we can understand Hank's use of irony when he says "Maybe if you give him a pack of cigarettes he'll tell you what he did with Jimmy Hoffa". The message Hank intends to convey is that this guy is not trustworthy and that he'll make up any story if he's given even a small incentive (even as small as one pack of cigarettes)—and, more specific to the current conversation, the message is that this guy is not the drug mastermind Heisenberg. Hank conveys this message by using irony. But the interesting part about this irony is that he's not saying something he believes is false; instead, he's presupposing something he believes is false. Specifically, "he'll tell you what he did with Jimmy Hoffa" presupposes that he (James Kilkelly) did something with Jimmy Hoffa (i.e., killed him and got rid of his body somewhere). Hank doesn't believe he actually did that; that much is obvious because (1) Hank doesn't really believe Kilkelly is a hardened criminal, and (2) in general it's widely known that people claiming to have killed Jimmy Hoffa are just telling "tall tales". So in this case, Hank is presupposing something he believes is false, in order to ironically show his attitude that anyone who believes that stuff is stupid. It's an interesting twist on irony (compare to the first example I gave, "Oh, pragmatics is so interesting", which is based on saying something false rather than presupposing something false).


With that background out of the way, have students think of other ways that an implicature might arise from presupposed, rather than entailed, content.


⟵ Alternatives and context
The pragmatics of presupposition ⟶

by Stephen Politzer-Ahles. Last modified on 2022-04-26. CC-BY-4.0.