3.4: Presupposition

In the previous section we discussed how the meaning of one sentence can entail the meaning of another sentence. Entailment is a very strong kind of inference. If we are sure that p is true, and we know that p entails q, then we can be equally sure that q is true. In this section we examine another kind of inference, that is, another type of meaning relation in which the utterance of one sentence seems to imply the truth of some other sentence. This type of inference, which is known as a presupposition, is extremely common in daily speech; it has been intensively studied but remains controversial and somewhat mysterious.

As a first approximation, let us define presupposition as information which is linguistically encoded as being part of the common ground at the time of utterance. The term common ground refers to everything that both the speaker and hearer know or believe, and know that they have in common. This would include knowledge about the world, such as the fact that (in our world) there is only one sun and one moon; knowledge that is observable in the speech situation, such as what the speaker is wearing or carrying; or facts that have been mentioned earlier in that same conversation (or discourse).

Speakers can choose to indicate, by the use of certain words or grammatical constructions, that a certain piece of information is part of the common ground. Consider the following example:

(10) “Take some more tea,” the March Hare said to Alice, very earnestly.
“I’ve had nothing yet,” Alice replied in an offended tone, “so I can’t take more.”⁵

By using the word more (in the sense which seems most likely in this context, i.e. as a synonym for additional) the March Hare implies that Alice has already had some tea, and that this knowledge is part of their common ground at that point in the conversation. The word or grammatical construction which indicates the presence of a presupposition is called a trigger; so in this case we can say that more “triggers” the presupposition that she has already had some tea. However, in this example the “presupposed” material is not in fact part of the common ground, because Alice has not yet had any tea. This is a case of presupposition failure, which we might define as an inappropriate use of a presupposition trigger to signal a presupposition which is not in fact part of the common ground at the time of utterance. Notice that Alice is offended — not only by the impoliteness of her hosts in not offering her tea in the first place, but also by the inappropriate use of the word more.

3.4.1 How to identify a presupposition

There is an important difference between entailment and presupposition with regard to how the nature of the speech act being performed affects the inference. If p entails q, then any speaker who states that p is true (e.g. I broke your jar) is committed to believing that q (e.g. your jar broke) is also true. However, a speaker who asks whether p is true (Did I break your jar?) or denies that p is true (I didn’t break your jar) makes no commitment concerning the truth value of q. In contrast, if p presupposes q, then the inference holds whether the speaker asserts, denies, or asks whether p is true. Notice that all of the three sentences in (11) imply that the vice president has falsified his dental records. (This presupposition is triggered by the word regret.)

(11) a. The vice president regrets that he falsified his dental records.
b. The vice president doesn’t regret that he falsified his dental records.
c. Does the vice president regret that he falsified his dental records?

In most cases, if a positive declarative sentence like (12a) triggers a certain presupposition, that presupposition will also be triggered by a “family” of related sentences (sentences based on the same propositional content) which includes negative assertions, questions, if-clauses and certain modalities.⁶ For example, (12a) presupposes that Susan has been dating an Albanian monk; this presupposition is triggered by the word stop. All of the other sentences in (12) trigger this same presupposition, as predicted.

(12) a. Susan has stopped dating that Albanian monk.
b. Susan has not stopped dating that Albanian monk.
c. Has Susan stopped dating that Albanian monk?
d. If Susan has stopped dating that Albanian monk, I might introduce her to my cousin.
e. Susan may have stopped dating that Albanian monk.

In addition to the presupposition mentioned above, (12a) also entails that Susan is not currently dating the Albanian monk; but this entailment is not shared by any of the other sentences in (12). This contrast shows us that presuppositions are preserved under negation, questioning, etc. while entailments are not.⁷

The “family of sentences” test is one of the most commonly used methods for distinguishing entailments from presuppositions. To offer another example, the statement The neighbor’s dog killed my cat presupposes that the speaker owned a cat, and entails that the cat is dead. If the statement is negated (The neighbor’s dog didn’t kill my cat) or questioned (Did the neighbor’s dog kill my cat?), the presupposition still holds but entailment does not.

Von Fintel & Matthewson (2008) describe another test for identifying presuppositions. They point out that if a presupposition is triggered which is not in fact part of the common ground, the hearer can appropriately object by saying something like, “Wait a minute, I didn’t know that!” This kind of challenge is not appropriate for information that is simply asserted, since speakers do not usually assert something which they believe that the hearer already knows:

A presupposition which is not in the common ground at the time of utterance can be challenged by ‘Hey, wait a minute!’ (or other similar responses). In contrast, an assertion which is not in the common ground cannot be challenged in this way. This is shown in [13]… The ‘Hey, wait a minute!’ test is the best way we know of to test for presuppositions in a fieldwork context. (von Fintel & Matthewson 2008)

(13) A: The mathematician who proved Goldbach’s Conjecture is a woman.
B₁: Hey, wait a minute. I had no idea that someone proved Goldbach’s Conjecture.
B₂: #Hey, wait a minute. I had no idea that that was a woman.

A fairly large number of presupposition triggers have been identified in English; a partial listing is presented below. For many of these it seems that translation equivalents in a number of other languages may trigger similar presuppositions, but so far there has been relatively little detailed study of presuppositions in languages other than English.⁸

a. Definite descriptions: the use of a definite singular noun phrase, such as Bertrand Russell’s famous example the King of France, presupposes that there is a uniquely identifiable individual in the situation under discussion that fits that description. Similarly, the use of a possessive phrase (e.g. my cat) presupposes the existence of the possessee (in this case, the existence of a cat belonging to the speaker).

b. Factive predicates (e.g. regret, aware, realize, know, be sorry that) are predicates that presuppose the truth of their complement clauses, as illustrated in (11) above.⁹

c. Implicative predicates: manage to presupposes try; forget to presupposes intend to; etc.

d. Aspectual predicates: stop and continue both presuppose that the event under discussion has been going on for some time, as illustrated in (12) above; resume presupposes that the event was going on but then stopped for some period of time; begin presupposes that the event was not occurring before.

e. Temporal clauses (14a–b) and restrictive relative clauses (14c) presuppose the truth of their subordinate clauses, while counterfactuals (14d) presuppose that their antecedent (if) clauses are false (see Chapter 19). Comparisons like (14e) presuppose that the relevant statement holds true for the object of comparison.

(14) a. Before I moved to Texas, I had never attended a rodeo.
(presupposes that the speaker moved to Texas)

b. While his wife was in the hospital, John worked a full 40 hour week.
(presupposes that John’s wife was in the hospital)

c. “I’m looking for the man who killed my father.”¹⁰
(presupposes that some man killed the speaker’s father)

d. If you had not written that letter, I would not have to fire you.
(presupposes that the hearer did write that letter)

e. Jimmy isn’t as unpredictably gauche as Billy.¹¹
(presupposes that Billy is unpredictably gauche)

The tests mentioned above seem to work for all of these types, but in other respects it seems that different kinds of presupposition have slightly different properties. This is one of the major challenges in analyzing presuppositions. We return in Chapter 8 to the issue of how to distinguish between different kinds of inference.

3.4.2 Accommodation: a repair strategy

Recall that we defined presuppositions as “information which is linguistically encoded as being part of the common ground at the time of utterance.” We crucially did not require that implied information actually be part of the common ground in order to count as a presupposition. We have already seen one outcome that may result from the use of presupposition triggers which do not accurately reflect the common ground at the time of utterance, namely presupposition failure (10 above). Another example of presupposition failure is provided in (15), taken from the 1939 movie The Wizard of Oz:

(15) Glinda: Are you a good witch or a bad witch?

Dorothy: Who, me? I’m not a witch at all. I’m Dorothy Gale, from Kansas.

Glinda: Well, is that the witch?

Dorothy: Who, Toto? Toto’s my dog.

Glinda: Well, I’m a little muddled. The Munchkins called me because a new witch has just dropped a house on the Wicked Witch of the East. And there’s the house, and here you are and that’s all that’s left of the Wicked Witch of the East. What the Munchkins want to know is, are you a good witch or a bad witch?

Glinda’s first question presupposes that one of the two specified alternatives (good witch vs. bad witch) is true of Dorothy, and both of these would entail that Dorothy is a witch. Dorothy rejects this presupposition quite vigorously. Glinda’s second question (Is that the witch?), and in particular her use of the definite article, presupposes that there is a uniquely identifiable witch in the context of the conversation. The fact that these false inferences are triggered by questions is a strong hint that they are presuppositions rather than entailments.

Glinda’s questions in this passage trigger presuppositions which Dorothy contests, because these inferences are not part of the common ground. However, presupposition failure is not the only possible outcome with such inferences. Another possibility is that the hearer, confronted with a mismatch between a presupposition trigger and the current common ground, may choose to accept the presupposition as if it were part of the common ground; in effect, to add it to the common ground. This is most likely to happen if the presupposed information is uncontroversial and consistent with all information that is already part of the common ground; something that the hearer would immediately accept if the speaker asserted it. For example, suppose I notice that you have not slept well and you explain by saying My cat got stuck on the roof last night; and suppose that I did not previously know you had a cat. Technically the presupposition triggered by the possessive phrase my cat is not part of the common ground, but I am very unlikely to object or to consider your statement in any way inappropriate. Instead, I will add to my model of the common ground the fact that you own a cat. This process is called accommodation.

It is not uncommon for speakers to encode new information as a presupposition, expecting it to be accommodated by the hearer. For this reason, definitions which state that presuppositions “must be mutually known or assumed by the speaker and addressee for the utterance to be considered appropriate in context” are misleading.¹² This fact has long been recognized in discussions of presupposition, as the following quotes illustrate:

I am asked by someone who I have just met, “Are you going to lunch?” I reply, “No, I’ve got to pick up my sister.” Here I seem to presuppose that I have a sister even though I do not assume that the speaker knows this. (Stalnaker 1974: 202).

It is quite natural to say to somebody… “My aunt’s cousin went to that concert,” when one knows perfectly well that the person one is talking to is very likely not even to know that one had an aunt, let alone know that one’s aunt had a cousin. So the supposition must be not that it is common knowledge but rather that it is non-controversial, in the sense that it is something that you would expect the hearer to take from you (if he does not already know). (Grice 1981: 190)

3.4.3 Pragmatic vs. semantic aspects of presupposition

Thus far we have treated presupposition primarily as a pragmatic issue. We defined it in terms of the common ground between a specific speaker and hearer at a particular moment, a pragmatic concept since it depends heavily on the context of the utterance and the identity of the speech act participants. Presupposition failure, where accommodation is not possible, causes the utterance to be pragmatically inappropriate or infelicitous.¹³ In contrast, we defined entailment in purely semantic terms: an entailment relation between two propositions must follow directly from the meanings of the propositions, and does not depend on the context of the utterance.

It turns out that presuppositions can have semantic effects as well. We have said that knowing the meaning (i.e. semantic content) of a sentence allows us to determine its truth value in any given situation. Now suppose a speaker utters (16a) in our modern world, where there is no King of France; or (16b) in a context where the individual John has no children; or (16c) in a context where John’s wife had not been in the hospital. Under those circumstances, the sentences would clearly not be true; but would we want to say that they are false? If they were false, then their denials should be true; but the negative statements in (17), if read with normal intonation, would be just as “un-true” as their positive counterparts in the contexts we have just described.

(16) a. The present King of France is bald.¹⁴
b. John’s children are very well-behaved.
c. While his wife was in the hospital, John worked a full 40 hour week.

(17) a. The present King of France is not bald.
b. John’s children are not very well-behaved.
c. While his wife was in the hospital, John did not work a full 40 hour week.

We have already noted that the presupposition failure triggered by such statements makes them pragmatically inappropriate; but examples like (16–17) show that, at least in some cases, presupposition failure can also make it difficult to assign the sentence a truth value. Some of the earliest discussions of presuppositions defined them in purely semantic, truth-conditional terms:¹⁵ “One sentence presupposes another just in case the latter must be true in order that the former have a truth value at all.”¹⁶

Under this definition, presupposition failure results in a truth-value “gap”, or indeterminacy. But there are other cases where presupposition failure does not seem to have this effect. For example, if (18a) were spoken in a context where the vice president had not falsified his dental records, or (18b) in a context where Susan had never dated an Albanian monk, these sentences would be pragmatically inappropriate because of the presupposition failure. But it also seems reasonable to say they are false (the vice president can’t regret something he never did; Susan can’t stop doing something she never did), and that their negative counterparts in (19) have at least one reading (or sense) which is true.

(18) a. The vice president regrets that he falsified his dental records.
b. Susan has stopped dating that Albanian monk.

(19) a. The vice president doesn’t regret that he falsified his dental records.
b. Susan has not stopped dating that Albanian monk.

However, there are various complications concerning the way negation gets interpreted in examples like (19). For example, intonation can affect the interpretation of the sentence. We will return to this issue in Chapter 8.

---

⁶ Cherchia & McConnell-Ginet (1990).

⁷ A more technical way of expressing this is to say that presuppositions project through the operators illustrated in (12), while entailments do not.

⁸ Exceptions to this generalization include Levinson & Annamalai (1992), Matthewson (2006), and Tonhauser et al. (2013).

⁹ Kiparsky & Kiparsky (1970).

¹⁰ Maddie Ross in the movie True Grit.

¹¹ Levinson (1983: 183).

¹² See for example http://en.Wikipedia.org/wiki/Presupposition.

¹³ We will give a more precise explanation of the term infelicitous in Chapter 10, as part of our discussion of speech acts.

¹⁴ Adapted from Russell (1905).

¹⁵ e.g. Frege (1892); Strawson (1950; 1952).

¹⁶ Stalnaker (1973: 447), summarizing the positions of Strawson and Frege. Stalnaker himself argued for a pragmatic analysis.