You are on page 1of 12

Presupposition projection with local contexts and CCPs

Clemens Mayr

Precedence in semantics, EGG school, Lagodekhi


mayr@zas.gwz-berlin.de

July 26, 2016

1 Admittance conditions and the projection problem

1.1 Admittance conditions


Recall that under the admittance condition view the presuppositions of a sentence S are those propositions that
the common ground of any context c (or rather its context set) must entail in order for S to be felicitous in
c (Stalnaker 1973, 1974, 1978, Karttunen 1974, Heim 1983). We say that c admits S iff its common ground
entails all of S’s presuppositions.

(1) a. Common ground of c =: {p : each participant x of c, x believes p}


b. Context set of c =: ∩{p : p ∈ common ground of c}
(Stalnaker 1978)

When uttering (2), the procedure in (3) takes place.

(2) Context: It is mutual shared belief that John used to smoke.


John stopped smoking.

(3) a. In all worlds in the common ground of c John smoked at some time in the past, i.e., the common
ground entails that John smoked at some time in the past.
b. c admits John didn’t stop smoking.
c. John didn’t stop smoking adds via its truth-conditions the information to c that John still smokes.

1.2 The projection problem under the admittance condition view


Heim (1983) can be seen as taking an insight from Gazdar (1979), namely that semantic evaluation happens
sequentially. However, she disagrees with him on two things: (i), presuppositions should be viewed as admit-
tance conditions, (ii) presupposition cancellation does not exist.

(4) The projection problem under admittance conditions


How to determine compositionally the admittance conditions of a complex sentence S based on the
admittance conditions of the parts of S?

This could be addressed in two possible ways (Stalnaker 1973, Karttunen 1974):

1
• How do we compositionally compute from the admittance conditions of the subparts of S the proposi-
tion corresponding to the admittance conditions of S?
• How can a context which entails only the proposition corresponding to the admittance condition of a
subpart of S nevertheless admit S?

A parallel question in another domain According to Heim (1983) one should ask the latter question. This
is a more general question as it can be asked about phenomena which traditionally are not thought to be related
to presuppositions (although they might actually be). For instance, third person pronouns seem to come with
certain admittance conditions:

(5) Context: The speaker tells the addressee that John, who is present, submitted a paper.
a. #She just submitted a paper to a journal.
b. He just submitted a paper to a journal.

We could now raise the projection problem for admittance conditions like in (5). For instance, why does (6)
not inherit the admittance conditions of the consequent—ignoring the possibility that the pronoun in (6) is not
of the same type as the one in (5), say because it is anaphoric whereas the latter is indexical?

(6) If I submit a paper with a co-author, she usually does the proof-reading.

Note that conventional implicature approaches have difficulty with (6): in a plugs-holes-filter approach the
conventional implicature associated with she in (6) could only be filtered if the antecedent entails it. This is
not the case here: the proposition that the speaker submits a paper with a co-author does not entail anything like
that there is a maximal salient female individual in the context. Under the cumulative approach the antecedent
would have to implicate that the speaker is not certain that there is a maximal salient female individual in the
context—not an obvious inference.
We know what goes on in (6). Uttering the antecedent changes the context by making a maximal co-
author salient, which can be picked up by a pronoun in the consequent. That this co-author is female can be
accommodated. So the real question, Heim maintains, should be:

(7) How does the context evolve intra- and intersententially so that the admittance conditions of a sentence
S are met?

2 Precursors to Heim (1983)

2.1 Stalnaker’s pragmatic picture


Stalnaker (1973) suggests that while semantically conjunction as in (8) is symmetric, pragmatically it is not.
That is, when uttering (8a), a speaker first asserts the first conjunct. Any proposition that is asserted becomes a
presupposition on this view as it is added to the common ground. Thereby once the second conjunct is asserted
its presupposition is already entailed by the new common ground. This is the reason why (8a) is felt to not
have a presupposition.
For (8b) to be assertable and in particular for the first conjunct the initial common ground must already
entail that John used to smoke. Therefore asserting the second conjunct is redundant. The sentence is felt to
be degraded.

(8) a. John has cancer, and he stopped smoking. 6 John used to smoke
b. #John stopped smoking, and he has cancer. John used to smoke

2
In other words, presupposition projection patterns are suggested to simply reflect the way assertions are per-
formed, namely sequentially.

2.2 Karttunen’s satisfaction theory


Karttunen (1974) suggests that admittance of a sentence S and S presupposing p are interdefinable:

(9) S presupposes p iff all contexts c admitting S entail p.

Similar to Stalnaker, Karttunen envisions that sentences and subsentences are added to the context/common
ground sequentially. He proposes that the incremented context c + Si can be seen as the local context for Si+1 .
This makes it possible to give a general definition of admittance as follows:

(10) Karttunen’s admittance conditions


a. A context c satisfies a presupposition p iff c entails p.
b. A context c admits a simple sentence φ iff c satisfies all of the presuppositions of φ.
c. A context c admits sentence φ iff for each subsentence φi of φ, the local context of φi admits φi .

The crucial question to address now becomes: what is the local context of a given subsentence Si ? Karttunen
suggests that local contexts have to be defined on a case by case basis.

(11) If If φ then ψ is uttered in context c, then c is the local context of φ and c + φ is the local context of ψ.

Consider (12).

(12) If the king has a son, the king’s son is bald. There is a king
6 The king has a son

According to the definitions above, context update proceeds as follows:

(13) a. c admits (12) iff each subsentence of (12) is admitted by its local context. (10c)
b. c admits (12) iff c admits The king has a son and c+The king has a son admits The king’s son
is bald. (12)
c. c admits The king has a son iff c entails that there is a king. (10a), (10b)
d. c+The king has a son admits The king’s son is bald iff c+The king has a son entails that the
king has a son. (10a), (10b)

It is clear that initial context c must be one in which there is a king in order for the antecedent to be admitted
by it (13c). Once this context is incremented by the antecedent it is automatically guaranteed that there is a
king with a son, and this therefore admits the consequent (13d). In other words, the initial context c need not
guarantee that there is a king with a son and thus need not admit the consequent on its own.
This raises the following question: is there a general rule for determining local contexts? (12) is pretty
much a stipulation. Karttunen more or less implicitly suggests that the sequence of context update and there-
fore also the computation of local contexts is done on a left to right basis.

3
3 Heim’s system based on context change potentials

3.1 The framework: context change potentials


3.1.1 What are CCPs again?

Heim (1983) suggests that a system of context change based on the one in (Heim 1982) has the notion of local
context already built in. Deciding on what the local context of a given sentence is is determined by the CCPs
of the logical operators used.
Heim assumes that the semantic content of a sentence φ is the context change potential (CCP) of φ—i.e.,
its ability to affect the context. The CCP of a simple sentence φ is the instruction to conjoin/intersect φ’s
(local) context c with the proposition φ or rather the proposition expressed by φ:

(14) Executing the CCP of φ on c


c+φ=c∩φ

The CCP of a complex sentence φ is determined on the basis of the CCPs of its parts.
She follows Karttunen’s view of admittance:

(15) Admittance
A context c admits a sentence φ iff c satisfies all of the presuppositions of φ.

3.1.2 CCPs and truth-conditional content

Heim argues that the truth-conditional content of any logical expression is derivable from its CCP. In particular,
she states that the connection between CCP and truth-conditional content of a sentence be that φ is true iff
adding φ to a true context outputs another true context:

(16) Truth of a context (to be modified)


A context c is true (in w) if w ∈ c.

(17) Truth of a sentence φ with CCPs


If c is true (in w) and c admits φ, φ is true (in w) with respect to c iff c + φ is true (in w).

Consider what this means for (18), as shown in the derivation in (19).

(18) John stopped smoking.

(19) a. Assume c is true, i.e., all the propositions in the common ground are true.
b. Assume c admits John stopped smoking, i.e., c entails that John used to smoke. (10a), (10b)
c. John stopped smoking is true with respect to c iff c+John stopped smoking is true. (17)
d. John stopped smoking is true with respect to c iff c∩that John stopped smoking is true. (14)
e. that John stopped smoking is true (definition of ∩)

3.1.3 CCPs and presupposition heritage properties

Heim further claims that the CCP of any logical expression determines its heritage properties. That is, once
one has the CCP both the truth-conditions and the projection properties follow, and it becomes unnecessary to
stipulate them independently of each other.

4
(20) Admittance conditions as definedness conditions of CCPs
c + φ is defined iff c admits φ.

3.2 Application of the theory


3.2.1 Presuppositions of conditional statements

Assume that the CCP for if φ, ψ is as in (21). Given (20) it follows that the execution of this CCP on c is only
defined if execution of the CCP of φ on c is defined—i.e., if c admits φ—and execution of the CCP of ψ on
c + φ is defined—i.e., if c + φ admits ψ.

(21) c + if φ, ψ = c \ ((c + φ) \ ((c + φ) + ψ))

In this way, the projection properties observed fall out from the CCPs associated with complex sentences.
Consider (22). As (23e) makes plain c does not need to satisfy the presupposition that John used to smoke.
Only the local context of the consequent—i.e., the initial context c incremented with the antecedent—must do
so. In (22) the nature of the antecedent guarantees that this is the case. Thus (22) as a whole does not have an
admittance condition, i.e., no presupposition.

(22) If John used to smoke, he stopped smoking. 6 John used to smoke

(23) a. c + (22) is defined iff c admits (22) (20)


b. c + (22) = c \ ((c + John used to smoke)\((c + John used to smoke) + he stopped smoking))
(21)
c. c admits (22) iff c admits John used to smoke and c + John used to smoke admits he stopped
smoking. (20), (23b)
d. c admits John used to smoke iff c satisfies all of the presuppositions of John used to smoke.
(15)
e. c + John used to smoke admits he stopped smoking iff c + John used to smoke satisfies all of
the presuppositions of he stopped smoking. (15)

Consider next (24). Given (25e), (24) has the admittance condition that c incremented with the proposition
that John has cancer entail that John used to smoke. This corresponds to a weak presupposition. This is so,
because the admittance condition says that only if c incremented with the proposition that John has cancer is a
true context, is it necessary that John used to smoke. I.e., we can paraphrase this as the conditional statement
that if John has cancer, he used to smoke. The prediction is thus parallel to that of Karttunen and Peters (1979).

(24) If John has cancer, he stopped smoking. John used to smoke

(25) a. c + (24) is defined iff c admits (24) (20)


b. c + (24) = c \ ((c + John has cancer)\((c + John has cancer) + he stopped smoking)) (21)
c. c admits (24) iff c admits John has cancer and c+John has cancer admits he stopped smoking.
(20), (25b)
d. c admits John has cancer iff c satisfies all of the presuppositions of John has cancer. (15)
e. c + John has cancer admits he stopped smoking iff c + John has cancer satisfies all of the
presuppositions of he stopped smoking. (15)

For (26), on the other hand, c must admit the antecedent and therefore satisfy its admittance condition that
John used to smoke. I.e., in (26) the admittance condition of the antecedent is inherited by the conditional as
a whole.

(26) If John stopped smoking, he has cancer. John used to smoke

5
(27) a. c + (26) is defined iff c admits (26) (20)
b. c + (26) = c \ ((c + John stopped smoking)\((c + John stopped smoking) + he has cancer))
(21)
c. c admits (26) iff c admits John stopped smoking and c + John stopped smoking admits he has
cancer. (20),(27b)
d. c admits John stopped smoking iff c satisfies all of the presuppositions of John stopped smok-
ing. (15)
e. c + John stopped smoking admits he has cancer iff c + John stopped smoking satisfies all of
the presuppositions of he has cancer. (15)

3.2.2 Other connectives

The CCP of negation is defined as in (28). One first executes the CCP of the simple sentence on c and
subtracts the result from c. Thus c must admit the simple sentence. This predicts negation to be a hole for
presuppositions.

(28) c + not φ = c \ (c + φ)

(29) John didn’t stop smoking. John used to smoke

The CCP for conjunction is as in (30). This makes the prediction that conjunctive sentences share their
admittance conditions with conditionals. (31) suggests that this correct.

(30) c + φ and ψ = (c + φ) + ψ

(31) a. John stopped smoking, and he has cancer. John used to smoke
b. John has cancer, and he stopped smoking. John used to smoke
c. John used to smoke, and he stopped. 6 John used to smoke
d. #John stopped smoking, and he used to smoke. John used to smoke

The CCP for disjunction is debatable. Heim does not discuss it. If it was similarly left to right based as
conjunction, something like (32) suggests itself (Karttunen 1974). This requires the initial context to satisfy
the presupposition of the first disjunct. I.e., it explains (33a). It also explains why (32c) does not have a
presupposition. The admittance condition of the second disjunct must be met in an incremented context c0
from which worlds in which John never smoked are removed from the initial context. I.e., the local context of
the second disjunct includes the negation of the first disjunct. As a consequence it is guaranteed in the local
context of the second disjunct that John used to smoke. But this also predicts a weak conditional presupposition
for (32b).

(32) c + φ or ψ = (c + φ) ∪ ((c \ (c + φ)) + ψ)

(33) a. Either John stopped smoking, or he doesn’t have cancer. John used to smoke
b. Either John doesn’t have cancer, or he stopped smoking. John used to smoke
c. Either John never smoked, or he stopped. 6 John used to smoke
d. Either John stopped smoking, or he never smoked. 6 John used to smoke

However, it seems that with disjunction filtering from right to left as in (33d) is possible. The CCP in (32)
cannot account for this. Thus possibly (34) is needed. The consequence would be that (33a) also gets a weak
conditional presupposition.

(34) c + φ or ψ = ((c \ (c + ψ)) + φ) ∪ ((c \ (c + φ)) + ψ)

6
3.2.3 Accommodation

What happens if the initial context c does not entail the presupposition of a sentence being uttered? Heim
suggests that the (local) context of the sentence is augmented so as to entail its presupposition. Then the
sentence’s CCPs is executed.

(35) Accommodation
Given sentence φ with presupposition p and context c which does not admit φ, proceed as follows:
a. c ∩ p,
b. c ∩ p + S.

There are two options: first in a case like (36), global accommodation is chosen. That is, the presupposition is
accommodated to the initial context c before any CCP is executed:

(36) Context: It is not common ground that John used to smoke.


John didn’t stop smoking.

(37) a. c + (36) is defined iff c admits (36) (20)


b. c ∩ John used to smoke = c 0 (35a)
c. c0 + John didn’t stop smoking = c0 \ (c0 + John stopped smoking) (28)
d. 0 0
c admits (36) iff c admits John stopped smoking (20), (37b)
e. 0 0
c admits John stopped smoking iff c satisfies all of the presuppositions of John stopped
smoking. (15)

In a case like (38), the presupposition that John used to smoke cannot be accommodated to the initial context
before executing the sentences CCP on it. After adding the because-sentence, the resulting context would be
inconsistent, i.e., false. In this case, local accommodation happens. That is, the presupposition is accommo-
dated to the local context of the sentence John stopped smoking but never to the initial context c:

(38) Context: It is not common ground that John didn’t use to smoke.
John didn’t stop smoking, because he never smoked.

(39) a. c + (38) is defined iff c admits (38) (20)


b. c + John didn’t stop smoking = c \ (c + John stopped smoking) (28)
c. c + John didn’t stop smoking =
c \ (c ∩ John used to smoke + John stopped smoking) (35a)
d. c admits (38) iff c ∩ John used to smoke admits John stopped smoking (20),(39c)
e. c admits John stopped smoking iff c ∩ John used to smoke satisfies all of the presuppositions
of John stopped smoking. (15)

The result of option two is what Gazdar (1979) achieved via his presupposition cancellation and what Kart-
tunen and Peters (1979) achieved by defining a second lexical entry for negation acting as a hole for presup-
positions.

3.3 Extension to quantificational examples


Similar to Gazdar’s 1979 account it is not immediately clear how to deal with presuppositions in the scope of a
quantifier. Since presupposition projection is seen as driven by context change one must address what context
change amounts to at the subsentential level.

7
3.3.1 File change semantics

Contexts as sets of sequence-world pairs Heim suggests that contexts should not be identified with sets
of propositions but rather as sets of sequence-world pairs. That is, the information accumulated in a context
is not all propositional but rather also keeps track of discourse referents, much like with file cards in a file.
For each discourse referent a file card is created. The cards contain information about that discourse referent.
A file is seen to be true iff there is a collection of individuals each corresponding to one of the cards and the
description on the cards is true of those individuals. Context change amounts to updating the information on
existing file cards and adding new ones. Upon encountering the subsentential unit in (40) the information is
French is added to file card number 5 if such a card exists. If it does not exist, it has to be established and the
information is also encoded on it.

(40) [ 5 is French ]

More technically, a context is identified with properties of sequences of individuals, i.e., a set of pairs hg, wi
with g a sequence of individuals also called an assignment function with the natural numbers as its domain
and the set of individuals as its range, and w a world. Each set of sequence-world pairs determines a unique
proposition:

(41) If c is a set of sequence-world pairs, the proposition determined by c is {w : ∃g.hg, wi ∈ c}.

Because of (41), identifying contexts with sets of sequence-world pairs is not disadvantageous when compared
to the identification of contexts with propositions. In particular, contexts can still be seen as true and false:

(42) Truth of a context


c is true in w iff ∃g.hg, wi ∈ c.

(43) Truth of a sentence φ with CCPs


If c is true (in w) and c admits φ, φ is true (in w) with respect to c iff c + φ is true (in w).

CCPs with variables The CCP for simple sentences with open variables in them looks as follows:

(44) c + [i . . . n φ] = c ∩ {hg, wi : φ(w)(g(i)) . . . (g(n)) = 1}

For (45) this means:

(45) [ t5 is French ]

(46) a. c + (45) is defined iff c admits (45) (20)


b. c + (45) = c ∩ {hg, wi : g(5) is French in w} (44)

Admittance conditions with open variables The admittance condition of a sentence like (47) with open
variable must be that the context c “entails” that g(5) has a sister. That is, c must be such that for every
hg, wi ∈ c it holds that g(5) has a sister in w.

(47) [ t5 likes his5 sister ]

We therefore define satisfaction by making reference to sequence-world pairs. The definition of admittance
stays the same:

(48) Admittance and satisfaction


a. A context c satisfies a presupposition i . . . nP iff ∀hg, wi ∈ c : P(w)(g(i)) . . . (g(n)) = 1.

8
b. A context c admits a sentence φ iff c satisfies all of the presuppositions of φ.

3.3.2 Presuppositions of quantified sentences

Presuppositions in the scopes of universal quantifiers The CCP for every is as in (49). We require that a
new file is created for the numeral i—that is, i must not have been in the domain of g.

(49) c + [Every i A, B] is defined iff ∀hg, wi ∈ c : i < Dom(g), if defined


c + [Every i A, B] = {hg, wi ∈ c : ∀a[hg [a/i], wi ∈ c + A → hg [a/i], wi ∈ c + A + B} (where g [a/i] is
just like g except that g [a/i] (i) = a)

For (50a) under LF (50b) we get as a result of the CCP in (49) the set of sequence-world pairs hg, wi where
for every individual g(2) it holds that if g(2) is a student in w, g(2) stopped smoking in w (51b). Moreover, the
scope [ 2 stopped smoking ] is only admitted in its local context if every sequence-world pair hg, wi where
g(2) is a student in w is such that g(2) also used to smoke in w. I.e., we get a universal presupposition: every
student used to smoke (50e).

(50) a. Every student stopped smoking. Every student used to smoke


b. [ every 2 [2 a student ] [ 2 stopped smoking ]]

(51) a. c + (50b) is defined iff c admits (50b) (20)


b. c + (50b) = {hg, wi ∈ c : ∀a[hg [a/2 ], wi ∈ c + [2 a student ] → hg [a/2 ], wi ∈ c + [2 a student ] +
[2 stopped smoking ]} (49)
= {hg, wi ∈ c : ∀a[hg [a/2], wi ∈ c ∩ {hg, wi : g(2) is a student in w} → hg [a/2 ], wi ∈ c ∩ {hg, wi :
g(2) is a student in w} ∩ {hg, wi : g(2) stopped smoking in w}} (44)
c. c admits (50b) iff c admits [ 2 a student ] and c + [ 2 a student ] admits [ 2 stopped smoking ]
(20), (51b)
d. c admits [ 2 a student ] iff c satisfies all presuppositions of [ 2 a student ] and c + [ 2 a student
] admits [ 2 stopped smoking ] iff c + [ 2 a student ] satisfies all presuppositions of [ 2 stopped
smoking ] (48b), (51c)
e. c + [ 2 a student ] admits [ 2 stopped smoking ] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2) is a student in
w} : g(2) used to smoke in w (48a), (51d)

Note that if we did not require in (49) that the initial context c must not contain a file card with number i on
it yet, we could not guarantee the correct truth-conditions. In particular, assume c = {hg, wi : g(2) is John
in w}. Then we would get (52), which only requires that every individual who is a student and John stopped
smoking.

(52) c + (50b) = {hg, wi ∈ {hg, wi : g(2) is John in w} : ∀a[hg [a/2], wi ∈ {hg, wi : g(2) is John in
w} ∩ {hg, wi : g(2) is a student in w} → hg [a/2], wi ∈ {hg, wi : g(2) is John in w} ∩ {hg, wi :
g(2) is a student in w} ∩ {hg, wi : g(2) stopped smoking in w}} (44)

Presuppositions in the restrictor Presuppositions in the restrictor of every do not as clearly project univer-
sally as those in the scope, as (53) shows:

(53) a. Every student who stopped smoking has cancer. ? Every student used to smoke
b. [ every 2 [2 a student who stopped smoking ] [ 2 has cancer ]]

However under the CCP given above and the somewhat simplified LF in (53b), the admittance conditions of
the restrictor are as in (54). That is, they require that every sequence-world pair hg, wi in c is such that g(2) is
a student who used to smoke in w. I.e., everyone is a student who used to smoke. This is much too strong.

9
(54) c admits [ 2 a student who stopped smoking ] iff ∀hg, wi ∈ c : g(2) is a student who used to smoke
in w (48a), (48b), (49)

The strong result in (54) can be weakened if we treat the relative clause as its own subsentential unit—as one
seems natural anyway. That is, the restrictor [ 2 a student ] would form part of the local context of the relative
clause. This is guaranteed by the CCP in (56).

(55) [ every 2 [[2 a student ] [ who2 stopped smoking ]] [ 2 has cancer ]]

(56) c + who2 φ = c ∩ {hg, wi : φ(w)(g(2)) = 1}

The resulting admittance conditions are as in (57), which only require that every sequence-world pair hg, wi
in c ∩ {h, g, wi : g(2) is a student in w} is such that g(2) used to smoke in w. This is thus a universal statement
about only the students.

(57) c ∩ {hg, wi : g(2) is a student in w} admits [ who2 stopped smoking ] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2)
is a student in w} : g(2) used to smoke in w (48a), (48b), (49), (56)

This, however, might still be too strong. Heim conjectures that there might be local accommodation going on.
The denotation would then be the set of sequence-world pairs hg, wi where for every individual g(2) it holds
that if g(2) is a student who used to smoke in w, g(2) stopped smoking in w (59a). The admittance condition
now requires that every sequence-world pair hg, wi in c ∩ {h, g, wi : g(2) is a student in w and g(2) used to
smoke in w} is such that g(2) used to smoke in w. This is tautological, i.e., no presupposition is predicted at
all.

(58) a. c+(55) = {hg, wi ∈ c : ∀a[hg [a/2], wi ∈ c+[2 a student ]∩{hg, wi : g(2) used to smoke in w} →
hg [a/2], wi ∈ c + [2 a student ] ∩ +[2 stopped smoking ]} (49), (35a), (57)
= {hg, wi ∈ c : ∀a[hg [a/2 ], wi ∈ c ∩ {hg, wi : g(2) is a student in w} ∩
{hg, wi : g(2) used to smoke in w} → hg [a/2], wi ∈ c∩{hg, wi : g(2) is a student in w}∩{hg, wi :
g(2) stopped smoking in w}} (44)
b. c ∩ {hg, wi : g(2) is a student in w and g(2) used to smoke in w} admits [ who2 stopped smoking
] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2) is a student in w and g(2) used to smoke in w} : g(2) used to
smoke in w (48a), (48b)

Negative quantifiers Negative existential quantifiers similarly seem to have a universal inference regarding
the presupposition in their scope.

(59) a. No student stopped smoking. Every student used to smoke


b. [ no 2 [2 a student ] [ 2 stopped smoking ]]

Heim’s account predicts this straightforwardly. The CCP is as in (60). That is the local context of the scope is
parallel to that of the scope of a universal quantifier.

(60) c + [No i A, B] is defined iff ∀hg, wi ∈ c : i < Dom(g), if defined


c + [No i A, B] = {hg, wi ∈ c : ¬∃a[hg [a/i], wi ∈ c + A ∧ hg [a/i], wi ∈ c + A + B}
(where g [a/i] is just like g except that g [a/i] (i) = a)

As a result, the admittance conditions for the scope are the same for (59) and the corresponding example with
every:

(61) c + [ 2 a student ] admits [ 2 stopped smoking ] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2) is a student in
w} : g(2) used to smoke in w (48a), (48b), (60)

10
However, given what said about local accommodation with respect to presuppositions in the restrictor above,
it appears that one should be able to weaken the presupposition also in the case of negative quantifiers. If we
accommodate locally in the scope we get as a denotation the set of sequence-world pairs hg, wi such that for
no individual g(2) it holds that g(2) is a student and g(2) used to smoke and stopped, (62a). This is compatible
with none of the students ever having smoked. The presupposition in (62b) requires that every student who
used to smoke used to do so. This is tautologous, i.e., no presupposition would be predicted anymore.

(62) a. c + (59b) = {hg, wi ∈ c : ¬∃a[hg [a/2], wi ∈ c + [2 a student ] ∧ hg [a/2], wi ∈ c + [2 a student ] ∩


{hg, wi : g(2) used to smoke in w} + [2 stopped smoking ]}
(60), (35a)
= {hg, wi ∈ c : ¬∃a[hg [a/2 ], wi ∈ c ∩ {hg, wi : g(2) is a student in w} ∧ hg [a/2 ], wi ∈ c ∩ {hg, wi :
g(2) is a student in w}∩{hg, wi : g(2) used to smoke in w}∩{hg, wi : g(2) stopped smoking in w}}
(44)
b. c ∩ {hg, wi : g(2) is a student in w} ∩ {hg, wi : g(2) used to smoke in w} admits [ 2 stopped
smoking ] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2) is a student in w} ∩ {hg, wi : g(2) used to smoke in
w} : g(2) used to smoke in w (48a), (48b)

There is also the option of accommodating in the restrictor—which is referred to as intermediate accommo-
dation (van der Sandt 1992, Beaver and Zeevat 2007). But it does not change the outcome in the present
case.

Indefinites But what about indefinites? Here it does not seem that a universal inference and thereby a
universal presupposition is warranted:

(63) a. A student stopped smoking. The student who stopped smoking used to smoke
b. [2 a student ] [ 2 stopped smoking ]

Given what has just been said about no and its relation to every, we would expect a universal presupposition
for (63a). Heim maintains that is actually not predicted by her given that indefinites are not quantifiers but
simply variables. That is, the CCP of (63a) is given (63b) simply the conjunction of two open sentences as in
(65). We already know how to interpret this. If the numeral 2 is not in the domain of g yet, we get adequate
truth-conditions as in (64a). However, even with this maneuver one gets a universal presupposition that every
student used to smoke, as (64b) shows.

(64) a. c + (64b) = c + [ 2 a student ] + [ 2 stopped smoking ] (44)


b. c ∩ {hg, wi : g(2) is a student in w} admits [ 2 stopped smoking ] iff ∀hg, wi ∈ c ∩ {hg, wi : g(2)
is a student in w} : g(2) used to smoke in w (48a), (48b)

For that reason, Heim suggests that there is accommodation as in (65a) occurring. This gives the presuppo-
sition in (65b) stating that if g(2) is a student who used to smoke, g(2) used to smoke. Together with the
denotation in (65a) it follows that g(2) used to smoke but not more.

(65) a. c + (65b) = c + [ 2 a student ] ∩ {hg, wi : g(2) used to smoke in w} +


[ 2 stopped smoking ] (44), (35a)
b. c ∩ {hg, wi : g(2) is a student in w} ∩ {hg, wi : g(2) used to smoke in w} admits [ 2 stopped
smoking ] iff ∀hg, wi ∈ c∩{hg, wi : g(2) is a student in w}∩{hg, wi : g(2) used to smoke in w} :
g(2) used to smoke in w (48a), (48b)

11
References
Beaver, David, and Henk Zeevat. 2007. Accommodation. In Oxford Handbook of Linguistic Interfaces, ed.
Gillian Ramchand and Charles Reiss. Oxford University Press.
Gazdar, Gerald. 1979. Pragmatics: Implicature, Pressuposition, and Logical Form. New York: Academic
Press.
Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation, University
of Massachusetts, Amherst.
Heim, Irene. 1983. On the projection problem for presuppositions. In Proceedings of WCCFL 2, ed. Daniel P.
Flickinger, 114–125. Stanford University, Stanford, California: CSLI Publications.
Karttunen, Lauri. 1974. Presupposition and linguistic context. Theoretical Linguistics 1:181–194.
Karttunen, Lauri, and Stanley Peters. 1979. Conventional implicature. Syntax and Semantics 11:1–56.
van der Sandt, Rob. 1992. Presupposition projection as anaphora resolution. Journal of Semantics 9:333–377.
Stalnaker, Robert C. 1973. Presuppostions. Journal of Philosophical Logic 447–457.
Stalnaker, Robert C. 1974. Pragmatic presuppositions. In Semantics and Philosophy, ed. M. Munitz and
D. Unger, 197–213. New York University Press.
Stalnaker, Robert C. 1978. Assertion. Syntax and Semantics 9.

12

You might also like