5 Presupposition

One of the things we noted in the last chapter is that the felicitous use of the definite article in English requires that the referent be familiar or unique or identifiable in some sense, although the particulars of the requirement were found to be very difficult to pin down. Closely related to this requirement is the fact that the use of the definite article seems to suggest an assumption of the existence of the referent; that is, to utter (154a) is to presuppose (154b):

One cannot felicitously utter (154a) without both the speaker and the hearer taking for granted that there is a King of France. Once again, then, we find that communication hinges on the mutual assumptions of the interlocutors regarding each other’s belief states.

You will not be surprised to learn that one of the arguments concerning presupposition has centered on whether it is a semantic or a pragmatic phenomenon, or both. This time, however, the argument is not based so much on how the fields of semantics and pragmatics are delimited as on how presupposition itself is defined.

5.1 Presupposition, Negation, and Entailment

If anything is asserted there is always an obvious presupposition that the simple or compound proper names used have a reference. If one therefore asserts ‘Kepler died in misery’, there is a presupposition that the name ‘Kepler’ designates something.

In short, to utter an assertion about Kepler is to presuppose that the term Kepler has a referent, that is, to presuppose that Kepler exists (or at least existed, before he died in misery). Nonetheless, this bit of meaning is not conveyed in the same way that “died in misery” is conveyed; at the very least, it’s apparent that the primary purpose of uttering Kepler died in misery would not be to convey that the name Kepler designates something, whereas it would indeed be to convey that the entity designated by this name died in misery. Frege moreover noted one other crucial property of presuppositions, which is that a presupposition carried by a given sentence will also be carried by its negation:

Here we have the negation of (154a), yet it continues to carry the presupposition in (154b) to the effect that there is a King of France. If the presupposition held the same status as the primary assertion of the sentence, Frege reasoned, the two should be similarly affected by negation, but clearly they are not: If (154a) means “the King of France is wise and there exists a King of France,” then the negation in (155) should mean “it is not the case that (a) the King of France is wise and (b) there exists a King of France,” which would be true if either the King of France is not wise or there is no King of France, but clearly this is not the case; one would not in general say that (155) is true if there is no King of France. Instead, the main assertion (the predication of wisdom) is essentially undone by negation, whereas both the positive and negative variants carry the same presupposition (of existence).

Frege believed that if a presupposition is false, the sentence containing it cannot have a truth value. Thus, if there is no King of France (for example, in the present real-world context), then neither (154a) nor (155) has any truth value at all. Later authors, however, have disputed this assessment. Thus, philosophers and linguists since Frege have wrestled with the nature of the presupposed bit of meaning – and relatedly, with the status of an utterance carrying a false presupposition. Notice that it is important to keep distinct the ramifications of negating the entire utterance and of finding the presupposition to be false. These are easy to confuse. For Frege, negating the utterance retains the presupposition, but if the presupposition is false, the utterance has no truth value.

For Russell (1905), on the other hand, the presupposition is part of the conveyed meaning of the utterance. Recall from Chapter 4 that in Russell’s view, a sentence with a definite NP subject is analyzed roughly as in (156):

Thus, the semantic meaning of the student arrived comes out to something like “there is a student, and there is no other entity that is a student, and that entity arrived.” Similarly, (154a) would be analyzed as “there is a King of France, and there is no other entity that is a King of France, and that entity is wise,” as shown in (157) (where “KoF” stands for “King of France”):

There is an important difference between this view and Frege’s: For Russell, if the presupposition is false, it does not render the utterance truth-valueless, as it does for Frege. Instead, to utter (154a) is to assert both the existence of the King of France and that he is wise; thus, the falsity of the presupposition entails the falsity of the utterance: If there is no King of France, then The King of France is wise is simply false. Thus, for Russell, the negation in (155) can indeed mean either that the King of France is unwise or that there is no King of France, as in (158):

The felicity of (158) has been taken by some as confirmation of Russell’s view; however, note that for the negation to be taken as a negation of the presupposition, it must be preceded by a claim that the King of France is wise, and followed up by a clarification such as that in (158), stating explicitly that it is the presupposition that is being negated. In the absence of such an addendum – that is, in the default case – (155) can only be taken as negating the King’s wisdom, not his existence. (More on this below.)

Strawson (1950) follows Frege in arguing that the statement expressed in (154a) is truth-valueless when (154a) is uttered in a context in which there is no present King of France. He points out, moreover, that the possibility of a truth value for this statement depends on when it is uttered; it has no truth value in the present context, but had one at other points in history. The truth conditions for the sentence depend on the context in which it is uttered because the expression the King of France is interpreted relative to that context; in that sense, its interpretation affects the truth conditions of the sentence just as much as the referent of the pronoun he affects the truth conditions of the sentence He is wise. Thus, if (154a) is uttered in 1840, its truth requires (among other things) the existence in 1840 of a King of France, whereas if it is uttered in 2012, its truth requires the existence in 2012 of a King of France. For Strawson, the truth of the presupposition in the current context is a necessary precondition for the truth or falsity of the utterance that presupposes it; given the lack of a King of France in 2012, the utterance cannot be true – at least not when spoken of the real world (as opposed to some fictional world).

Notice, however, that in this sense Strawson has shifted the analysis of presupposition in a pragmatic direction (not that he would have phrased it in those terms), in which the context is important. A truly semantic account of presupposition would frame the notion in purely semantic terms. As Levinson (1983) puts it, semantic presupposition in its simplest form can be defined as follows:

This is a definition that can hold of sentences as opposed to utterances (i.e., it can be the case that a sentence A will only be true in a situation in which B is true, etc.). However, as Levinson points out, in a bivalent (two-valued) system of logic this entails that B is always true, since the truth of A entails the truth of B and the falsity of A entails the falsity of B, and A (regardless of its content) is necessarily either true or false. Since it would seem that merely presupposing a proposition should not entail its truth (independent of the truth of the sentence that presupposes it), a semantic view of presupposition such as that sketched in (159) would seem to require that we abandon the concept of a two-valued logical system – that is, a system with the values “true” and “false” for every proposition, and no other – and accept instead a system with at least one intermediate value of “neither true nor false.” This is exactly the type of system required by the Frege/Strawson account of presupposition, in which the falsity of the presupposed proposition entails that the presupposing utterance is neither true nor false. The semantic system of presupposition we arrive at, then, is as follows:

The relationship between the (a) and (b) clauses in both formulations reflects the hallmark of presupposition, which is constancy under negation. Constancy under negation gives us our clearest test for distinguishing presuppositions from entailments. If A entails B, negating A does not retain the entailment, but rather leaves the status of B entirely unspecified:

This is a standard case of scalar entailment, as discussed in Chapters 2 and 3. Here we see that (161a) entails (161b), since it’s impossible to own three crowns without owning two. The negation of (161a), shown in (161c), does not share this entailment; if the King of France does not own three crowns, it is entirely possible that he owns none at all. In contrast, the presupposition in (161d) is shared by both (161a) and (161c). Thus, the presupposition is constant under negation, whereas the entailment disappears under negation – which means that constancy under negation can distinguish between entailments and presuppositions. Moreover, this test does not depend in any way on the choice between a Strawsonian and a Russellian view of presupposition, since in both systems a sentence and its negation share presuppositions.

The fact that a sentence and its negation share presuppositions also makes it difficult to respond to a presuppositional query with a simple yes or no without being taken to accept the presupposition, as in (162):

The addressee cannot simply answer yes or no to the question in (162a) without in some sense agreeing to the presupposition that they have smoked in the past, since both (162b) and (162c) share this presupposition. Instead, they would have to explicitly deny this presupposition using a form along the lines of that seen above in (158), as in (162d). The fact that a simple no cannot be used to deny the presupposition provides support for a Strawsonian view over a Russellian view, as noted above. More generally, as we will see below, the fact that the presupposition can be cancelled as in (162d) has been used as evidence that a purely semantic account is insufficient. It also gives us an additional way to distinguish between presupposition and entailment, since entailments cannot be cancelled.

Notice that the presupposition is preserved not only in the negation of the presupposing sentence, but also in cases of questioning and suspension, as in (163b–d):

As noted in Chierchia and McConnell-Ginet (2000), these variants form a family of expressions which preserve the presupposition that there exists a King of France, despite the negation (b), questioning (c), or suspension (d) of the proposition “the King of France is wise.”

Such existence presuppositions can, by their nature, subtly affect a person’s beliefs about the world, and even about their own experience. Loftus and Zanni (1975) conducted an experiment in which subjects were shown a short film of a car crash. Afterward, they were asked one of two questions – either Did you see a broken headlight? or Did you see the broken headlight? The question with the definite article – the one that presupposed the existence of a broken headlight – elicited significantly more yes responses than the question with the indefinite article, even when no broken headlight actually was present. This and related experiments suggest that the use of presuppositional questions can cause people to believe they have seen or otherwise experienced something that they did not experience at all – a finding with important ramifications for courtroom questioning, advertising, and many other types of interaction.

That presupposition and entailment are distinct is further evidenced by the fact that although a presupposition is typically also entailed, it is not the case that an entailment is typically also presupposed. Consider (164):

In (a), it’s my wife that’s been kidnapped both entails and presupposes (b). We can see that (b) is presupposed by noting that the negation in (c) shares the presupposition; we can see that it is entailed by noting that it is impossible for the presupposing sentence in (a) to be true without (b) also being true. In (d), on the other hand, (b) is entailed (i.e., if (d) is true, (b) is necessarily true), but it is clearly not presupposed, as can be seen by noting that the negation in (e) conveys nothing about whether (b) is true, that is, about whether anyone has been kidnapped at all.

Finally, as noted above, the fact that presuppositions can sometimes be cancelled distinguishes them from entailments, which cannot be cancelled. Thus, whereas one can follow up an utterance like I haven’t stopped smoking (which presupposes “I used to smoke”) with I never did smoke! in (162d), cancelling the presupposition, one cannot very well follow up an utterance like My wife has been kidnapped (which entails “someone’s been kidnapped”) with an attempt to cancel the entailment; that is, My wife has been kidnapped; nobody has been! is simply gibberish. This difference follows directly from the fact that a sentence and its negation share presuppositions but not entailments. While we can negate the first clause and get the perfectly reasonable My wife hasn’t been kidnapped; nobody has been!, the negated first clause in this case, My wife hasn’t been kidnapped, does not retain the entailment “someone’s been kidnapped,” so the second clause does not count as cancelling an entailment of the first. The reason (162d) works as a cancellation of a presupposition is precisely because the first clause, despite being the negation of I have stopped smoking, retains the presupposition. For this reason, it cannot be the case that all presuppositions are entailed: Whereas a positive sentence such as (162b) entails its presupposition, a negated sentence such as (162c) cannot, since if the presupposition were entailed it could not be cancelled.

5.2 Presupposition Triggers

Although we have thus far focused on a relatively small set of examples, there are a great many expressions and constructions that give rise to presuppositions. These are termed presupposition triggers, and they can be classed into a variety of categories, a handful of which are illustrated in the examples in (165)–(169):

These are the types of cases that we have been discussing so far, and which were the focus of the early discussions of Frege, Russell, and Strawson: Use of a definite description presupposes the existence of its referent. Thus, in (165a), use of the definite NP the check that has been the Burberry emblem since 1924 presupposes that such a check exists and has been the Burberry emblem since 1924, while in (165b) the use of the italicized NPs presupposes both that the Shire exists and that Bilbo had a remarkable disappearance and unexpected return.

Factive verbs (Kiparsky and Kiparsky 1971) are verbs that take a sentential complement and presuppose that complement. Thus, in (166a), the verb notice serves as a trigger indicating that its sentential complement (that Frodo also showed signs of good “preservation”) is presupposed, while in (166b) the verb recall likewise serves as a trigger indicating that its sentential complement (that that breather year saved his life) is presupposed. To notice something presupposes that it is true (you cannot notice something that is not the case); similarly, to recall something presupposes that it occurred.

A change-of-state verb, as its name suggests, indicates a shift from one state to another, and therefore presupposes that the moved-from state has held at some point in the past. Thus, in (167a) use of the verb stop presupposes that hands-on work has been devalued in the past (you can’t stop devaluing it unless you have at some point devalued it), while in (167b) stating that the gate opened presupposes that it was closed to begin with.

Much as change-of-state verbs presuppose the moved-from state, iteratives indicate repetition of some past action or state, and thus presuppose that that past action occurred or that the past state held. In (168a), we see two examples of this: For Mary and the children to return to New York presupposes that they were in New York in the past, and for someone to be alone again, he must have been alone in the past. Similarly, in (168b), for Zizi to bring the basin again presupposes that she brought the basin at some point in the past.

Clefts (which will be discussed in more detail in Chapter 7) are a group of presuppositional syntactic structures that come in several forms, of which the most common are the it-cleft (also known simply as a cleft), the wh-cleft (also known as a pseudo-cleft), and the inverted wh-cleft (also known as a reverse pseudo-cleft):

As we will see in more detail in Chapter 7, use of a cleft structure focuses one constituent (in (170a–c), a short-circuit) while presupposing the rest of the propositional content of the utterance (in (170a–c), that something caused the power failure). Thus, in (169a), the italicized it-cleft presupposes that someone taught him to dress, and in (169b), the italicized inverted wh-cleft presupposes that something gave the speaker the courage and confidence to go to New York.

All of the above categories of presupposition can be verified by checking for constancy under negation:

These sentences are negations of the examples in (165)–(169), and share their presuppositions; (171a) presupposes the existence of the check as Burberry’s emblem, (171b) that the Shire exists and that Bilbo had a remarkable disappearance and unexpected return, (171c) that Frodo showed signs of good preservation, (171d) that that breather year saved his life, and so on. Notice that similar expressions lacking the presupposition will have negations that also lack the presupposition:

In these cases, the presupposition of the original – that the breather year saved his life, that Zizi had brought basin previously, and that someone taught him how to dress – is absent in these instances, both in the positive and the negated variants. Moreover, notice again that what is entailed (rather than presupposed) by an utterance is not entailed (nor presupposed) by its negation; for example, (172a) entails that Hamilton believed something, but (172b) does not; (173a) entails that the girl Zizi brought something, but (173b) does not; and (174a) entails that he learned to dress, but (174b) does not. Thus, the test of constancy under negation can distinguish both between presupposition and its absence and between presupposition and entailment.

5.3 The Projection Problem

Researchers studying presupposition soon found themselves faced with what has come to be known as the projection problem for presupposition – that is, the question of what accounts for the difference between cases in which a presupposition carried by an embedded expression “percolates up” to the embedding expression and cases in which it does not. Consider (175):

In (175a–b), we see that the use of the factive verb realize presupposes the complement; thus, (a) presupposes he’s the King of France, while (b) presupposes the Burberry emblem is attractive. But there’s another presupposition that also survives. In each of the two embedded sentences, there’s an existential presupposition – in (a), that there is a King of France, and in (b), that there is a Burberry emblem. Each of these continues to be presupposed by the full sentence in (a) and (b), respectively. Karttunen (1973) introduced the term holes for linguistic expressions and operators that allow the presuppositions of their component expressions to pass through to the larger expression. As we’ve seen above, negation is a hole, since a negated sentence retains the presuppositions of the positive variant; factive verbs are holes as well, as we see in (175a–b).

In (175c), on the other hand, there is no presupposition that the King of France exists. Here the verb think is what’s called a propositional-attitude verb, in that it expresses the subject’s attitude toward some proposition (here, the proposition expressed in the embedded clause). In the context of such verbs, the presupposition of the embedded clause may vanish from the larger sentence, as it does here. Karttunen uses the term plugs for expressions that prevent inheritance of a presupposition in this way, including propositional-attitude verbs, as in (175a), and verbs of saying, as in John said he’s the King of France. However, the matter is not nearly that straightforward. Notice that in (175d), for example, the use of the verb thinks does not eliminate the presupposition; (175d) presupposes that there exists a Burberry emblem. It turns out that it’s very difficult, if not impossible, to find a class of linguistic expressions which consistently “plug” presuppositions – that is, which consistently prevent presuppositions from projecting upward to the containing sentence. Karttunen himself acknowledges that all plugs leak; others (e.g., Levinson 1983) have argued that it’s unlikely that plugs exist at all.

Finally, Karttunen uses the term filters for connectives such as if … then, illustrated in (176a–b), and or, illustrated in (176c–d), which allow the presupposition to pass through to the larger construction in some instances but not others:

Here, (a) does not presuppose that the basin exists, whereas (b) does. (Note that these count as if … then sentences even though the word then doesn’t explicitly appear.) Similarly, (c) does not presuppose that Zizi forgot the basin, but (d) does. Karttunen (1973) describes the conditions under which filters operate: For example, the filtering seen with if … then, as in (a), is contingent on the antecedent (if … ) clause entailing what is presupposed by the consequent clause (he would have … ); that is, when the antecedent clause entails the presupposition of the consequent clause, that presupposition is filtered out of the larger sentence. Thus, because the girl Zizi had brought a basin in the antecedent of (a) entails the existence of a basin, the presupposition of the consequent (that the basin exists) is filtered out of the larger sentence. In (b), the antecedent does not have this entailment, and so the presupposition survives and the larger sentence presupposes the existence of the basin.

In the case of disjunction, as in (176c–d), Karttunen states that the presuppositions of the two component clauses will be inherited by the larger sentence unless the presupposition of the second component clause is entailed by the negation of the first. Thus, in (c), because the presupposition of the second clause (that she forgot the basin) is entailed by the negation of the first clause (Zizi did not remember the basin), the presupposition does not project upward to the larger sentence. In (d), on the other hand, the negation of the first clause (Zizi is not crying because she doesn’t feel well) does not entail the presupposition of the second (that she forgot the basin), and so the presupposition is preserved.

Although these conditions account for the data, they’re unsatisfying because they’re stipulative: They simply assert when the presupposition will be filtered out in a sentence containing one of these connectives, without really explaining why this should be the case. It’s as though these rather complex properties were arbitrarily associated with the lexical items if and and, without being due to more general principles. But that seems unlikely; it’s hard to imagine, for example, another language with a word that has the same meaning and use as English if except that in a sentence like (176a) the presupposition of the second component clause is retained by the whole sentence. (See Chierchia and McConnell-Ginet 2000 for a similar argument.)

5.4 Defeasibility

The projection problem has proved a difficult one for semanticists to get a firm handle on, all the more so because pragmatic factors appear to be crucial: Part of the reason for the difference between (175c) and (175d), for example, may lie in the difference between the two types of belief in question. For an individual to think he’s the King of France (regardless of whether there is a King of France) renders his judgment suspect; hence we are less likely to adopt the presuppositions carried by his reported beliefs. On the other hand, for an individual to admire a certain kind of emblem is not at all unusual, and we are therefore more willing to let the presupposition stand. More generally, we will see that presupposition is sensitive to contextual factors; and as we saw in Chapter 2, this sensitivity to contextual factors gives rise to the defeasibility of certain pragmatic phenomena – that is, their ability to be defeated, or cancelled, in the right context. Presupposition, it turns out, shares this quintessentially pragmatic property of being defeasible in certain contexts. We have already seen several examples of presuppositions being defeated, such as in the case of propositional-attitude verbs (e.g., (175c)) or in sentences with certain connectives (e.g., (176a, c)). We will now consider additional circumstances in which a presupposition may be defeated.

For example, you will recall that one of the hallmarks of presupposition is constancy under negation. However, we have also seen that this constancy is not quite as, well, constant as one might like. Thus, we find paradigms such as that in (177):

As we have seen, the litmus test for whether (177a) presupposes (178) is whether (177b) shares the presupposition – that is, whether (177a) and (177b) both convey an assumption of (178). If only (177a) conveys that the King of France exists and (177b) does not, then (178) is an entailment, not a presupposition, of (177a). If both convey (178), it’s a presupposition of both.

The fly in the ointment comes in the form of (177c), in which we see that the presupposition is defeasible – that is, it can be cancelled, just as a conversational implicature can be cancelled. So the question is: Is the presupposition constant under negation or not? In view of (177c), it would appear not. However, notice the rather marked nature of (177c): It is clear that this utterance can only be used in a very specific sort of context. Horn (1985) characterizes cases such as (177c) as instances of metalinguistic negation, in which, rather than negating the primary assertion (as with garden-variety negations), the speaker uses negation to object to virtually any aspect of the utterance at all, including for example the pronunciation of individual words (I didn’t eat the toMAHto, I ate the toMAYto) or, in this case, the presupposition. For this reason, metalinguistic negation requires an appropriate prior utterance: In the case of a metalinguistically negated presupposition, it requires a prior utterance that carries the presupposition being negated. For example, (177c) is felicitous only when it is preceded by a claim that the King of France is wise, in which case the metalinguistic negation serves to reject the presupposition. In the absence of that prior utterance, the utterance of (177c) serves no purpose, and it is infelicitous. Moreover, as noted above, the initial clause of (177c) (The King of France is not wise) will not be taken as negating the presupposition unless it is immediately followed by an explicit denial of that presupposition (there is no King of France); in any other context, it will be taken as denying the King’s wisdom, not his existence. That is, responding to (177a) with (177b) will not convey that there is no King of France; it is necessary to utter (177c), or something similar, for that effect.

In addition, this metalinguistic cancellation of the presupposition happens only via the negation of the presupposing utterance; for example, it is impossible to cancel the presupposition of (177a) using the positive variant in (179):

Even if the prior utterance asserts the negative variant, the presupposition remains the same, and so the negative variant is still required for the cancellation:

Of course, one could reasonably object that the problem with (179) lies with the infelicity of attributing the property of wisdom to an entity that one is simultaneously asserting does not exist. Recall that the King of France is wise not only presupposes, but also entails, the existence of the King of France; thus, (179) expresses a contradiction. Compare (179) with (181), in which the first clause of B’s utterance does not entail the existence of the King of France:

Here there is no infelicity, despite the absence of a negation in the first clause of B’s utterance. Since this clause does the same job as would have been accomplished by an explicit negation of A’s utterance – the cancellation of A’s presupposition – it is not necessary to have the explicit negation of A’s utterance. Indeed, in this case it would not be necessary to have the explicit negation of the presupposition (there is no King of France), either; B’s initial clause has the advantage of clearly denying the King’s existence, whereas the mere negation of A’s utterance (The King of France is not wise) requires the follow-up negating the presupposition, since otherwise, as noted above, the utterance will be taken as denying the King’s wisdom, not his existence. Most relevantly for our purposes, (181) does not appear to be a metalinguistic negation, yet it succeeds in cancelling the presupposition of the King’s existence. This would again seem to argue for presupposition having a pragmatic component.

Finally, recall also that at many points in history, uttering (177a) would have been unproblematic. As we noted at the beginning of the chapter, if someone had uttered The King of France is wise at one of the many points in history when France had a king, and if the king at that time was in fact wise, it would have been true. To some extent this returns us to the issue discussed in Chapter 3, in which establishing the referent of an NP – a clearly pragmatic matter – must take place before truth values can be fixed. So this is not an issue that’s entirely unique to presupposition. It does, however, again suggest a pragmatic component, with contextual factors (such as the year in which the sentence is uttered) playing a crucial role.

What, then, does this all mean for presupposition? It has been argued (e.g., Levinson 1983) that the existence of examples like (177c) argues against a purely semantic account of presupposition, since under a semantic account negation would have to be systematically ambiguous between a type of negation that preserves the presupposition and a type that negates it; such an ambiguity is of course unlikely, and there is really no independent evidence for it. And we have seen above in our discussion of metalinguistic negation a number of reasons to believe that presupposition has a pragmatic aspect. Moreover, metalinguistic negation isn’t the only way to cancel a presupposition. For one thing, as we saw above in our discussion of (176), they may be filtered out in some circumstances. For another, they may be suspended (Horn 1972), as in 182:

In these cases, the material in the antecedent of the conditional suspends the presupposition: It is no longer necessarily the case that John ever smoked in (182a), or that anyone taught him how to dress in (182b), or that anything gave the speaker the courage and confidence to go to New York in (182c). Notice that this isn’t quite the same thing as cancellation; in cancellation, the presupposition is removed entirely, whereas in suspension, the speaker explicitly declines to take a stand either way on the status of the presupposition. Nonetheless, the ability to suspend presuppositions in certain cases further argues for their having a pragmatic aspect.

There are other ways to cancel a presupposition as well; consider for example (183):

In (183a), there is a clear presupposition that Hatfield was called to the Pacific Theater in World War II, and indeed negating the sentence preserves the presupposition:

In (183b), however, there is no presupposition that Gandhi was awarded the Peace Prize, despite the fact that the relevant (italicized) portion of the utterance has the same basic structure as that in (183a). Likewise, in (183c), there is no presupposition that Professor Stroup heard his song in the celebrations being referred to. Presumably the difference lies in our mutual knowledge concerning what sorts of events can preempt other sorts of events – for example, that death can preempt the receiving of a Peace Prize or the hearing of a song; thus, compare the presupposition in (183a) with the absence of a presupposition in (185):

Here there is no presupposition that the individual under discussion was expelled, because people who have finished their degree cannot thereafter be expelled.

Similarly, consider again the examples from our discussion of the projection problem in (175) above, repeated here as (186):

Recall from our discussion of filters and holes that whereas factive verbs like realize in (186a–b) allow the presupposition of the King’s existence to percolate up from the embedded clause to the larger sentence, propositional-attitude verbs such as think in (186c–d) are inconsistent in whether they allow the presupposition through; the survival of the presupposition may depend more on pragmatic factors – such the plausibility of the presupposition in view of the interlocutors’ mutual knowledge – than on semantic factors such as the semantic category of the verb. That is, (186c) lacks the presupposition for reasons having to do with real-world knowledge and, perhaps, the perceived reliability of the subject’s referent, whereas (186d) retains the presupposition. (It’s possible that the difference in syntactic position of the phrase in question also plays a role, but notice that John thinks his coat carries the Burberry emblem shares the existential presupposition of (186b).)

The negation in (187a) does not retain the presupposition that we have been devaluing hands-on work, in view of the immediate denial of that presupposition; on the other hand, a similar attempt at denial in (187b) is distinctly odd. In (187a), the change-of-state verb stop is what Abusch (2002) terms a soft trigger in that the presupposition it triggers is easily cancelled, whereas the cleft construction in (187b) is a hard trigger whose presupposition is much less easily cancelled. Following Abusch, and building on a suggestion made by Bill Ladusaw, Abbott (2006) proposes that the difference between hard and soft triggers is whether the presupposition is detachable – that is, whether the only reason for saying it “in that way” would be to convey the presupposition. For example, notice that the speaker of (188a) could have easily chosen instead to utter (188b):

The only thing conveyed by the use of the cleft in (188a) beyond what is conveyed by the non-cleft variant in (188b) is precisely the presupposition that something gave me the courage and confidence to go to New York. This, Abbott argues, is the reason it’s odd to choose the cleft variant in the absence of the presupposition, and in turn is the reason why it is infelicitous to attempt to cancel the presupposition associated with the use of the cleft, as in (187b): It is inappropriate for the speaker to specifically choose a construction that serves only to convey the presupposition when they intend immediately to deny the presupposition. In contrast, the word stop in (187a) conveys more than simply the existence of a prior state (i.e., the presupposed notion that we have been devaluing hands-on work); it also conveys the cessation of that state. Since there is no way to convey the cessation of a state without presupposing its prior existence, the presupposition is nondetachable. Thus, because there are reasons to choose the word stop other than to convey the presupposition, it is not infelicitous to cancel the presupposition. Nonetheless, this does not appear to account for all hard/soft triggers, and there is more work to be done on this topic (see Abbott 2006 for further discussion).

In short, if presuppositions can be cancelled in some cases, and fail to arise for contextual or world-knowledge reasons in other cases, it would seem that presupposition is, after all, a pragmatic phenomenon. If it were purely semantic, we would not expect to be able to cancel the presupposition without contradiction, and we would not expect that contextual factors would affect whether or not a presupposition arises.

5.5 Presupposition as Common Ground

In contrast to a holes-and-filters type of approach to presupposition, which emphasizes the role of individual lexical items and constructions and offers rules for how they affect the heritability of presupposition, Stalnaker (1974, 1978) and others following him have emphasized the pragmatic aspects of presupposition, arguing in particular that presuppositions are part of the common ground of the discourse, that is, what is considered part of the interlocutors’ shared background information (or at least what is taken as uncontroversial; see the discussion of accommodation in the next section). On this view, if the presupposition does not hold, the utterance is inappropriate; to utter The King of France is wise in a world that contains no King of France is communicatively pointless and therefore bizarre.

If presupposition is defined in terms of the common ground shared by the speaker and hearer, many of the problems with purely semantic approaches disappear. In most of the problematic cases we’ve encountered, the primary problem has been that mutual knowledge, context, and the information presented in the utterance itself – all pragmatic aspects of the discourse – can override the presupposition. Moreover, there are also differences in the strengths of various presuppositions, which again suggests that pragmatic principles are involved:

As we’ve seen, (189a) strongly presupposes the king’s existence, whereas (189b) is entirely neutral on the matter. Example (189c) seems to fall somewhere between the two, and unlike (189a), it seems straightforwardly false in a world lacking a King of France. Note that its negation does not seem to assume that there is a King of France, which argues for the king’s existence being an entailment rather than a presupposition in this case:

And of course in (189d) the King of France is taken as being no more real than the Emperor of Venus, which Joey might have selected instead as inspiration for his costume.

Part of the issue is no doubt the fact that the NP the King of France can be used in reference to a particular entity or merely as a description of a property; that is, (191a) can be interpreted in a way parallel to either (191b) or (191c):

In (191b), the uncle is being equated with Barack Obama, and for this reason such a sentence is called an equative. In (191c), the uncle is not being equated with a particular entity; rather, the speaker is predicating a property of the uncle; hence such a sentence is called predicational. It stands to reason that equative uses tend to be presuppositional in a way that predicational uses are not; thus, a reading of (191a) as parallel with (191b) – that is, an equative use that identifies the uncle with an entity known as the King of France – will presuppose the existence of the King of France, whereas a reading of (191a) as parallel with (191c) – that is, a predicational use that attributes to the uncle the property of French king-hood – will not. Because (191a) is ambiguous between the two readings, its presuppositionality is less clear than that of (189a). Thus, the presuppositionality (or lack thereof) of a definite NP is in part due to how the speaker (is believed to have) intended the NP to be taken – as equative or predicational; here again we see pragmatic factors intruding on the interpretation of potentially presuppositional expressions. Donnellan (1966) makes a similar point; his referential/attributive distinction (see discussion in Chapter 4) was posited largely to counter the views of Russell and Strawson concerning the meaning of definite descriptions, by noting that the referential and attributive uses of definite NPs affect truth values differently in the case of presupposition failure. (You’ll recall from that discussion that Donnellan argued that Smith’s murderer is insane could be true of some person who was not actually Smith’s murderer, if that’s the person the speaker intended to refer to and if the referred-to person were in fact insane.) Donnellan considers uses such as that in (191c) to be non-referential, arguing that Is de Gaulle the King of France? does not seem to presuppose a King of France, whereas Is the King of France de Gaulle? does.

As Stalnaker (1974) points out, a pragmatic approach that views presuppositions as part of the shared background assumptions of the interlocutors gives us a ready account of such cases as (176) above, repeated here as (192):

Rather than requiring a set of rules for when filtering will and won’t occur with various logical connectives, as suggested by Karttunen (1973), Stalnaker argues that in such cases, the set of shared assumptions that form the background against which the second clause of the sentence is understood is richer than that against which the first clause is understood, precisely because it has been updated with the information in the first clause. Simply put, in (192a), the context for he would have washed his face in the basin includes if the girl Zizi had brought a basin, which preempts the presupposition. In contrast, in (192b), the first clause contains no information to counter the existence of a basin, so the presupposition of the second clause survives. Examples (192c–d) are treated similarly: In (192c) the first clause explicitly raises the possibility that Zizi remembered the basin, and since that forms part of the shared background for the second clause, the presupposition that would otherwise arise (that she forgot the basin) fails to arise; in (192d) the first clause contains nothing to counter this presupposition, so again the presupposition survives. In short, this account allows information already present in the common ground to cancel the presupposition, and this common ground can include information presented in an earlier clause within the same sentence.

Note that such an approach isn’t limited to pragmatic accounts; Heim (1983b) and Chierchia and McConnell-Ginet (2000), for example, present semantic accounts that similarly allow for the continual updating of the discourse model – and see Chapter 9 for further discussion of such “dynamic” approaches to meaning. These approaches may also be usefully compared with that of Gazdar (1979a, 1979b), who takes sentences to give rise to a set of potential presuppositions which are only actualized if they do not conflict with the sentence’s entailments and conversational implicatures. While Stalnaker’s approach and the other dynamic approaches are essentially linear (with earlier-added information affecting presuppositions associated with later-added information), Gazdar’s is essentially hierarchical (with entailments and conversational implicatures essentially outranking presuppositions); nonetheless, what all of the approaches share is that each offers a mechanism for allowing information within the same sentence to preempt a presupposition without the need for rules attached to specific lexical items.

If we take a pragmatic view of presuppositions as background information, one way to look at an utterance is to distinguish between the backgrounded, presupposed portion and the new, informative portion. This approach divides the information encoded by an utterance into a presupposition and focus. This division is very easy to see in the case of clefts. Consider the examples of clefts in (169), repeated here along with, for each, the cleft in question, its presupposition, and its focus:

In (193a) the focus his mother indicates that “his mother” is the “someone” who taught him how to dress; similarly, in (193b), that provides the “something” that “gave me the courage and confidence to go to New York.” Since that is anaphoric, the prior linguistic context provides its referent – that is, what it is that gave the speaker the courage and confidence to go to New York (specifically, the community support that she had found in Chicago).

Clefts are among a number of expressions that are felicitous only if an appropriate proposition is presupposed in the context. This means that (193a) will only be felicitous in a context in which it is presupposed that someone taught him how to dress, and (193b) will only be felicitous in a context in which it is presupposed that something gave the speaker the courage and confidence to go to New York.

5.6 Accommodation

Unfortunately, a view of presuppositions as belonging to the common ground also runs into trouble. Consider the following:

This is the beginning of an article, so the first NP – Robert Earl Keen, the Texas-based songwriter and performer – represents new information, as evidenced by the appositive that explains who Keen is. Nonetheless, Keen’s existence is presupposed, as evidenced both by the definite article and the fact that negating the sentence would retain this presupposition:

Similarly, our steel player’s brother represents information that is not in the common ground – the reader cannot be expected to know that the steel player has a brother (or even, for that matter, that there is a steel player) – yet this NP, too, clearly represents an entity whose existence is presupposed, as again evidenced both by its definiteness and by the fact that negating the sentence retains the presupposition:

How is it that both the steel player and the steel player’s brother can be treated as presupposed? They do not constitute part of the common ground, if being part of the common ground means being mutually known or taken for granted. But they are somehow plausible enough that the reader is willing to grant their existence once the writer has treated them as presupposed. Nonetheless, not just any entity can be treated this way; consider the following:

In (197a), as in (194), the addressee is likely to unquestioningly grant the existence of the speaker’s brother, even if this brother’s existence was previously unknown, presumably because it is uncontroversial for people to have brothers. Similarly, in (197b) the existence of the motorcycle is plausible enough to be presupposed uncontroversially. In (197c), the horse might not get by without comment, but any number of contextual matters might figure in here; in Lexington, Kentucky, this presupposition might pass unnoticed, whereas in downtown Chicago, it’s likely to be met with something like Hold it; you have a horse? Plausibility of ownership isn’t the only factor at work, however; in (197d), the presupposition associated with my apples seems distinctly odd, despite the fact that it is entirely uncontroversial for people to own apples. And my almonds in (197e) is even worse. Granted, it is less common (in America) for someone to have almonds at any given moment than for them to have apples, but it’s still not particularly unusual, and toasting almonds isn’t a particularly odd activity. Finally, in (197f), there’s no reason for an addressee not to grant the existence of a sculpture, yet the utterance is odd. In this case, the peculiarity may lie in the singular; I dusted my sculptures sounds a bit better, though not entirely good. One could argue that most people who own a sculpture own more than one, so it’s the uniqueness associated with the definite that’s infelicitous. Notice that if the speaker and addressee both know of the particular apples/almonds/sculpture in question (if, say, the speaker and hearer were together when the item in question was purchased earlier that morning), all three utterances become fine. The question, then, is only under what circumstances a previously unknown entity can be treated as presupposed in the discourse.

The need for a theory of pragmatics to be able to account for such cases has long been recognized. One early and very influential account (though not the first) is Lewis (1979). Lewis’s account likens a discourse to a game of baseball. Just as the baseball game’s current score is a direct function of an earlier state in combination with whatever plays have occurred since then, the current state of the discourse model is a direct function of an earlier state in combination with whatever discourse moves – utterances – have occurred since then. Thus, an utterance like (198) can be used to add a steel player and that player’s brother to the discourse model:

Once I’ve added this utterance to the discourse, the steel player and the brother can thereafter be presupposed; I have, in a sense, changed the “score” of the discourse model, in the same way that I can change the score of a baseball game by making a home run.

But there’s another way in which I can change the conversational score: I can treat these entities as though they’re already presupposed. In this way, conversation differs from baseball: You can’t change the score by running to home without first hitting the ball – that is, by behaving as though you had hit the ball. But in conversation, we can do something very much along those lines. If a conversational “player” treats something as presupposed, it counts as presupposed, regardless of its previous status in the discourse. In this sense, the conversational “score” is whatever we behave as though it is – regardless of the facts of the prior discourse. Thus, if the speaker mentions our steel player’s brother, it is as though the steel player’s brother had been previously known, hence is presupposed in the discourse, despite the fact that he actually constitutes new information. In this case, we say that the addressee accommodates the presupposition.

However, as noted above, accommodation isn’t unlimited in its ability to treat new information as presupposed. There are limits to what a hearer will permit a speaker to slip into the conversation as presupposed. Saying that you phoned your brother is one thing; saying you toasted your almonds is quite another. Notice, moreover, that it’s not just existential presuppositions that are accommodated:

Separating the clefts here into presupposition and focus, as we did in (193) above, we find that in (199a), the presupposition is “Mankiewicz felt compelled to defend his actions at some time,” and the focus is only decades later. But there is nothing in the prior context to indicate that Mankiewicz felt compelled to defend his actions at some time. Nonetheless, the reader obligingly accommodates the presupposition. Similarly, in (199b) the presupposition is “the part of Massachusetts that curves into the ocean like an arm was named Cape Cod at some time,” and the focus is in the 1600s. Again, however, there is nothing in the prior context to make salient or accessible the notion that Cape Cod received its name at some time. Indeed, not only is the presupposition easily accommodated in this context, but this cleft could easily stand as the first sentence of a book (minus the last, anaphoric phrase), as illustrated in (200):

Thus, not only can accommodation occur when the context does not support the presupposition; it can occur in the relative absence of any shared context. You might (and should) object that in (200) the reader can be assumed to know that Cape Cod exists, and that it must have been named at some point, so it’s not quite true to say there’s no shared context – only that there is no shared salient context. But no such shared knowledge can be presumed for (201):

Here, the addressee is clearly not expected to have prior knowledge of the presupposed event. Prince (1978, 1986) calls such examples informative-presupposition clefts, since their very purpose is to inform the addressee of the information contained in the so-called “presupposition.” So if we are going to consider such instances to in fact constitute presuppositions – and note that they do pass the negation test – it cannot be the case that presupposition requires the presupposed material to be part of the common ground.

One way of thinking about presupposition that may avoid this difficulty is to think of the information expressed in an utterance as being either asserted or not asserted, with the unasserted material being presupposed; this is the approach taken by Abbott (2000, 2008). Thus, in (199a) the proposition “Mankiewicz felt compelled to defend his actions” is not what is being asserted; what’s being asserted is that this defense didn’t occur until decades later. Similarly, in (199b) the author is not asserting the proposition “the part of Massachusetts that curves into the ocean like an arm was named Cape Cod, after this delicious fish,” but rather that this happened in the 1600s. The fact that these presuppositions aren’t present in the common ground, then, is not relevant; rather, the cleft allows them to be presupposed in the sense of not being asserted, the better to focus on the material that is in fact being asserted. This may not entirely resolve the situation of informative-presupposition clefts, since the presupposition in (201) seems to have as a primary purpose the assertion of the presupposed material. However, given that there need not be a correspondence between what is asserted and what is new (Abbott 2008), and that there are limits on how much can be asserted in a given sentence (Abbott 2000), presupposition gives the speaker the option of detaching newness from assertion: A presupposition such as that in (201) allows the writer to simultaneously convey both “it was ten years ago today” and “young Irwin Vamplew was bopped on the head …” – both of which constitute new information – by asserting the former and presupposing the latter.

Note also that the type of speech act may be relevant; for example, an existential presupposition that would not pass muster in a declarative can sometimes get by in a request:

Although (202a) might raise a hearer’s eyebrow, it’s far more acceptable (in the absence of a mutually known shovel) than (202b). There is an element of plausibility that’s relevant – it’s clearly more acceptable to make reference to a previously unknown brother than to a previously unknown shovel – but there appear to be other issues at work as well. And the issue is clearly related to the unresolved questions raised in Chapter 4 concerning when an entity can and cannot be referred to with a definite NP. Thus, one might be tempted to argue for a familiarity-based account of definiteness under which unfamiliar entities such as that represented by the italicized NP in (203) below are simply accommodated, or one might alternatively be tempted to argue for a uniqueness-based account of definiteness under which non-unique entities such as that represented by the italicized NP in (204) are accommodated:

As noted in Chapter 4, the italicized definite in (203) represents information that is unfamiliar but uniquely identifiable, whereas the italicized definite in (204) represents information that is not uniquely identifiable but is familiar. And in fact the referent of a definite NP may be neither:

Here, it’s entirely possible that the guy in question is neither familiar (in that the hearer needn’t have any prior knowledge of this individual) nor uniquely identifiable (in that there may have been two guys sitting next to the speaker, one on each side).

The possibility of accommodation, then, makes the problem of definiteness even thornier: Any account of definiteness that allows for accommodation will demand that the circumstances under which accommodation is possible be spelled out precisely; otherwise the entire theory becomes vacuous. That is, the theory cannot simply be of the form “these are the conditions under which an NP may be definite – but NPs can also be definite if they don’t satisfy these conditions, in which case they are accommodated.” It is clear that the pragmatics of definiteness, presupposition, and accommodation are interrelated, and that any theory that purports to account for one will need to take the others into consideration as well.

5.7 Summary

In this chapter we discussed the phenomenon of presupposition. We began with semantic accounts, comparing accounts under which failure of presupposition renders the utterance false with those under which failure of presupposition renders the utterance without a truth value. From a consideration of what happens to the utterance when the presupposition is false, we moved to a consideration of what happens to the presupposition when the utterance is false, and found that constancy under negation served as a reliable test for presupposition. We examined the behavior of a wide range of presupposition triggers, that is, classes of expression that reliably give rise to presuppositions. We then considered the projection problem – the question of when a presupposition does and does not project upward from a subpart of an utterance to the utterance as a whole. This, along with related questions concerning the cancellability of presupposition, led us to conclude that a purely semantic analysis cannot fully account for presuppositional phenomena, and we then took up the possibility of a pragmatic account under which presuppositions constitute the shared background of an utterance. This definition, too, proved inadequate in light of the common phenomenon of accommodation, in which material that does not constitute part of the common ground is treated as nonetheless presupposed. We compared a variety of situations in which particular presuppositions can and cannot be accommodated, and proposed explanations for these differences. Finally, we considered the necessary relationship between theories of presupposition and theories of definiteness.

5.8 Exercises and Discussion Questions

1. Photocopy a paragraph of 10 lines or more from any book you wish, and underline each expression that seems presuppositional. On a separate sheet, list the presuppositions and test them for constancy under negation.

2. Use a truth table to show that if Russell is right in his characterization of presupposition (given in (157)), then he is also right that the falsity of the presupposition entails the falsity of the entire utterance.

3. Find six examples of presupposition in advertisements. Describe and explain each example, showing how it contributes to the goals of the advertiser.

4. Consider the following examples:

(i) The soup has thawed.

(ii) The soup has not thawed.

(iii) The soup has not thawed; it was never frozen!

(iv) The soup has thawed, if it was ever frozen.

Give the two presuppositions found in (i), and for each of the examples in (ii)–(iv), explain what effect (if any) the modification has on what is presupposed, using the terms and concepts from this chapter.

5. Abbott (2006) offers the following example to illustrate the difference between presupposition and conventional implicature:

(i) Even the King of France is bald. (Abbott 2006, example 12)

This is true if there’s a King of France and he’s bald, regardless of whether or not he’s the least likely person to be bald. However, it cannot be true if there’s no King of France. Explain how these two facts distinguish what is presupposed in (i) from what is conventionally implicated.

6. Conduct a web search to create a corpus of 50 naturally occurring instances of presupposition, including 10 of each of the five types of presupposition trigger described in the text.

7. Iteratives (e.g., again) are treated here and in much of the pragmatics literature as presuppositional, in that Zizi brought the basin again and Zizi did not bring the basin again both seem to assume that Zizi brought the basin previously. Others, however, have argued that this is a conventional implicature, not a presupposition. On what sort of evidence would the difference depend? How would you argue for one view over the other?

8. Presuppositions are easiest to deal with when the utterance under discussion is a declarative. What would Russell and Strawson say about an example such as (i)?

(i) Has Harriet stopped smoking?

How, in turn, might a pragmatic account deal with such an utterance?

9. We observed with respect to (182a), repeated below, that it’s possible to suspend a presupposition:

(i) John has stopped smoking, if he ever did smoke. (= (182a))

Explain why this mechanism appears to be available in cases like (i) but unavailable in cases like (ii):

(ii) #I realize that I broke the vase, if in fact I did.

Try to formulate a rule that will distinguish between these two categories of attempted suspensions.

10. The following sentence contains a presuppositional expression:

(i) Charlie regrets that he is tall.

Tell which category of presupposition trigger is involved here, and list five more members of this category other than those presented in the text.

11. Find eight examples of accommodation in written material. For each, explain what is being presupposed and why the reader is willing to accommodate the presupposition.

12. Recall Donnellan’s argument (from Chapter 4) that on the attributive reading, if nobody murdered Smith, the utterance Smith’s murderer is insane cannot be true. Relate this to the claim that if there is no King of France, the utterance The King of France is wise cannot be true. Does this mean that the difference between Donnellan’s referential and attributive categories boils down to the difference between NPs that are and are not presupposed? Support your answer.

(172)	a. Hamilton believes that that breather year saved his life. b. Hamilton does not believe that that breather year saved his life.
(173)	a. The girl Zizi brought the basin. b. The girl Zizi did not bring the basin.
(174)	a. His mother taught him how to dress. b. His mother did not teach him how to dress.

(159)	A sentence A semantically presupposes another sentence B if:
	(a) in all situations where A is true, B is true, (b) in all situations where A is false, B is true. (Levinson 1983)

(160)	If sentence A presupposes proposition B, then:
	(a) in all situations where A is true, B is true, (b) in all situations where A is false, B is true, (c) in a situation where B is true, A may be either true or false, (d) in all situations where B is false, A is neither true nor false.

(165)	Definite descriptions:
	a. The sharpest words I heard him use were “not nice,” speaking of counterfeiters who have co-opted the check that has been the Burberry emblem since 1924. (Collins 2009) b. Bilbo was very rich and very peculiar, and had been the wonder of the Shire for sixty years, ever since his remarkable disappearance and unexpected return. (Tolkien 1954)

(167)	Change-of-state verbs:
	a. “We need to stop devaluing hands-on work,” she said before the engines drowned her out. (Sullivan 2009) b. Frodo and Sam stopped dead, but Pippin walked on a few paces. The gate opened and three huge dogs came pelting out into the lane, and dashed towards the travellers, barking fiercely. (Tolkien 1954)

(177)	a. The King of France is wise. b. The King of France is not wise. c. The King of France is not wise; there is no King of France!
(178)	There is a King of France.

(197)	What did you do after I saw you this morning?
	a. I phoned my brother. b. I washed my motorcycle. c. I fed my horse. d. ?I peeled my apples. e. ?I toasted my almonds. f. ?I dusted my sculpture.

(203)	In her talk, Baldwin introduced the notion that syntactic structure is derivable from pragmatic principles. (= Chapter 4, (114a))
(204)	[To spouse, in a room with three equally salient windows] It’s hot in here. Could you please open the window? (= Chapter 4, (115a))