Presupposition

PHIL 43916
November 7, 2012

1. What is presupposition? ..................................................................................................1

2. Presupposition accommodation .......................................................................................2
3. The projection problem ...................................................................................................3
4. Complications with context sets ......................................................................................5

1. WHAT IS PRESUPPOSITION?
Presupposition is best introduced by example. Consider the following sentence:

Jane knows that Bob is a USC fan.

This sentence could not be felicitously uttered in a context in which it was a matter of
dispute whether Bob was a USC fan.

This might, however, seem less than surprising; after all, the sentence entails that Bob is
a USC fan and, in general, one can’t simply assert claims which trivially entail
propositions which are at issue in a conversation.

The interesting thing is that this relationship between the sentence and the proposition
survives various transformations of our original sentence, including:

negation: Jane doesn’t know that Bob is a USC fan.

question: Does Jane know that Bob is a USC fan?
conditional: If Jane knows that Bob is a USC fan, then she will avoid him at all
costs.

None of these entail that Bob is a USC fan; but, still, none could be felicitously uttered in
a context in which it was a matter of dispute whether Bob was a USC fan.

This might remind you of some of the examples of pragmatic implicatures we have
discussed; e.g., the example of the letter of recommendation writer, or ‘You are the cream
in my coffee.’ But note that those examples are highly context-dependent, and that they
don’t survive the sorts of transformations exemplified above.

Here are a few more examples:

 Jane stopped drinking wine for breakfast. (presupposes: Jane used to drink wine
for breakfast)

It was Bob that organized the cheating ring. (presupposes: Someone organized
the cheating ring.)

Bob’s children are obnoxious. (presupposes: Bob has children)

Smith returned to the scene of the crime. (presupposes: Smith had been at the
crime scene before)

2. PRESUPPOSITION ACCOMMODATION

This is all pretty suggestive of some pervasive linguistic phenomenon; but it would be
good to have a more precise characterization of what that phenomenon is. A natural
suggestion is something like the following. Let’s use the term common ground as a name
for the collection of propositions that all parties to a conversation take for granted. Then
we might say:

S presupposes a proposition P iff S can only be felicitously uttered in a

conversation if P is part of the common ground of the conversation.

The problem — sometimes called the problem of informative presuppositions — is that

this simply does not fit the cases listed above. The following conversation is perfectly
felicitous even if the first speaker does not know that Bob has children:

“Bob seems really exhausted.”

“Well, part of the problem is that Bob’s children are really obnoxious.”

Here the second sentence does not rely on this proposition being part of the common
ground; on the contrary, it appears to be an attempt to add this proposition to the
common ground. This is sometimes called presupposition accommodation.

This phenomenon poses an immediate problem for our characterization of presupposition;

and this is bad, because we want a clear description of the phenomenon for which we will
try to give a theoretical explanation. A plausible suggestion is something like the
following:

S presupposes a proposition P iff S can only be felicitously uttered in a

conversation if either (i) P is part of the common ground of the conversation or
(ii) participants in the conversation would be willing to add P the context set
without objection.

2
(Strictly, we should distinguish between what a sentence presupposes, and what a
particular utterance of a sentence presupposes. We won’t worry about this distinction for
present purposes.)

We can thus think of presupposition accommodation has having an effect on the common
ground. We can make this explicit by (following Stalnaker and others) thinking of the
common ground as the set of worlds consistent with everything that everyone in the
conversation takes for granted (and takes the others to take for granted). We can call this
world the context set. Then we might make the following prediction:

If S presupposes P, then if S can be felicitously uttered, one of the following must be the
case:
(i) P is true in every world in the context set
(ii) participants in the conversation are willing to add P to the common ground,
in which case the utterance of S eliminates all worlds in which P is not true
from the context set

Hence, whether or not we have a case of accommodation, the result of uttering S is that
all the worlds in the context set are worlds in which P is true.

An interesting possible application of this model: it explains the otherwise puzzling fact
that both sentences which predicate racial slurs of individuals, and the negations of those
sentences, are offensive. This might be explained in terms of the fact that negations of
sentences carry the presuppositions of those sentences, plus the claim that slurs are
offensive because of the presuppositions they carry.

This is a rough gloss of the standard model for thinking about presupposition. We’ll now
go on to explore some further complications in understanding the phenomenon.

3. THE PROJECTION PROBLEM

It is worth recalling again the contrast between presupposition and cases like ‘You are the
cream in my coffee.’ It appears that, in at least many cases, a given utterance’s
presuppositions are largely independent of the context; ‘Bob’s children are obnoxious’
seems to presuppose that Bob has children no matter what the purposes of the
conversation. Further, speakers seem to be able to recognize this fact just on the basis of
their linguistic knowledge.

This suggests — by argument parallel to our original argument for the existence of a
compositional semantics — that we ought to be able to give something like a
compositional treatment of the presuppositions of sentences. That is, we ought to be able
to explain the presuppositions of complex sentences in terms of the presuppositions of
their parts. For, if there were no such explanation, it would be mysterious how speakers

3
could figure out the presuppositions of the indefinitely many complex sentences for which
they have this ability.

The problem of computing the presuppositions of complex sentences on the basis of the
presuppositions of their parts is called the projection problem for presuppositions.

An attractively simple attempted solution to the projection problem is:

All complex sentences inherit all of the presuppositions of their parts.

This fits the cases discussed earlier, but does not fit every case. Consider, for example, the
following complex sentences, both of which contain the simple sentence ‘Jane knows that
Bob is a USC fan’:

Jane doesn’t know that Bob is a USC fan.

Sam said that Jane knows that Bob is a USC fan.

As we have seen, the first (like the simple sentence) carries the presupposition that Bob is
a USC fan. The second, however, does not. It seems that while negation lets the
presuppositions of the simple sentence through to the complex sentence, ‘said’ somehow
blocks the presuppositions of the simple sentence. Terms which are like negation are
called holes (because they let presuppositions through), and terms which are like ‘said’
are called plugs.

If all we had were plugs and holes, things would be pretty simple. But we don’t; we also
have filters: expressions which sometimes do, and sometimes do not, let presuppositions
through. If-then statements are paradigm cases here. Consider the difference between the
following two sentences:

If Jane has met Bob, then Jane knows that he is a USC fan.
If Bob is a USC fan, then Jane knows that he is a USC fan.

The first carries the presupposition of its consequent — that Bob is a USC fan. But the
second does not. This is the kind of thing that an answer to the projection problem must
explain.

In the case of if-then sentences, a standard view is that something like the following
generalization is correct:

A sentence ‘if S1 then S2’ carries the presuppositions of S2 unless those

presuppositions are entailed by S1.

4
‘Jane met Bob’ does not entail that Bob is a USC fan, which is why the first sentence
above carries that presupposition; ‘Bob is a USC fan’, however, does (trivially) entail that
Bob is a USC fan, which is why the presupposition is blocked in the second sentence
above.

(This rule for ‘if-then’, by the way, doesn’t seem to fit very well with the view, mentioned
above, that the offensiveness of racial/ethnic slurs is explained by the presuppositions
they carry.)

So far, this is just a story about one particular filter: ‘if-then.’ To give a satisfactory
solution to the projection problem, it seems that we would have to, at least, sort all
expressions in to plugs, holes, and filters and, for each filter, give generalizations like the
one we just gave for ‘if-then.’

But even if we could do this, it is plausible that we would still lack a real understanding
of the phenomenon of presupposition. This is because we would simply be tacking on to
our semantic theory extra rules to capture this phenomenon. What we should really want
is an account of why these expressions have the presupposition-related properties that
they have. Ideally, this would be a theory which explains why expressions with a certain
intension (or character) give rise to the presuppositions that they give rise to.

One possible stumbling block for this explanatory program is the apparent existence of
expressions with the same intension, but different presupposition-related properties. One
example is ‘but’ and ‘and.’ Many have thought that these share an intension, and yet the
following seem to differ in their presuppositions:

She is poor and honest.

She is poor but honest.

4. COMPLICATIONS WITH CONTEXT SETS

A different sort of complication in the standard story about presupposition and context
sets can be brought out by considering uses of ‘too.’ (The following is based on some
examples from Kripke, ‘Presupposition and anaphora.’) This expression seems to carry
presuppositions. For consider a situation in which we are debating whether anyone is
going to go to the bar. If I then say,

I’m going to the bar, too.

My utterance is infelicitous in just the way that utterances typically are which presuppose
something which is a matter of dispute. The presupposition carried by ‘too’ in this case
seems to be that people besides me are going to the bar.

5
But now consider a different case. Suppose you ask me what I’m going to do tonight, and
I say:

I’m going to North Dining Hall for dinner tonight, too.

Suppose that North Dining Hall has not been mentioned yet in the conversation. Here my
utterance seems bizarre. But it is puzzling why it should seem bizarre. What it
presupposes is that people other than me are going to to North Dining Hall for dinner
tonight. But this surely is a part of the common ground of our conversation; both of us
believe this to be true, and know that the other person believes it to be true. (And even if
we did not believe this, it is obvious that it is true, so at the least we should have
accommodation; but we don’t in this case.)

‘Too’ seems to require us to complicate our story about context sets above. It seems that
we have to distinguish between an active and a passive context set — where the former
includes the information on which we are actively focusing, and the latter also includes
information which we jointly take for granted, but aren’t focusing on right now. ‘Too’
seems to place constraints on the active context set, and not just the passive context set.

(Note, by the way, that the active context set is not just a matter of the utterances made
in our conversation; if I utter the above sentence about NDH in a situation in which
we’ve both just seen a bunch of students hurrying into North Dining Hall for dinner, then
the utterance seems fine even if none of us say anything about North Dining Hall.)

But this sort of phenomenon might also force us to complicate our view of context sets as
sets of worlds. Consider ‘again’, which is similar in some ways to ‘too’ in requiring
something of the active context set rather than merely of the passive context set. (Even if
we know that Bob, like most people, eats dinner every night, this fact alone is not enough
to make an utterance of ‘Bob is eating dinner again’ felicitous.)

Suppose we’re in a situation in which part of the active context set is the proposition that
everyone walked to the dining hall earlier today. Now suppose, looking at Bob sitting in a
chair, I say without further stage-setting:

Bob’s left foot is in front of his right foot again.

This seems bizarre. But why? The sentence seems to presuppose that at some earlier
time, Bob’s left foot was in front of his right foot. But given our assumption that the
proposition that everyone walked to the dining hall earlier is a salient topic and part of
the common ground, all of the worlds in the active context set are worlds in which at
some earlier time Bob’s left foot was in front of his right. This should suffice to make the
above utterance felicitous; but it doesn’t.

6
The problems here are analogous to the problems we encountered with attitude
ascriptions; to give an adequate treatment of the context set (whether active or passive)
we seem to need a more fine-grained conception of the information which is comprises the
common ground than that which is given by sets of worlds.

For more on this sort of issue, see Soames, ‘Kripke on presupposition and anaphora’,
available at:

http://www-bcf.usc.edu/~soames/forthcoming_papers/Kripke_on_Presupposition.pdf