Counterfactual Conditionals and Focus...
Transcript of Counterfactual Conditionals and Focus...
1
Counterfactual Conditionals
and Focus Crosslinguistically
Toshiyuki Ogihara
University of Washington
1. Introduction
This paper discusses counterfactual conditionals like (1a-b) in which the time indicated by the
tense morpheme(s) and the time described by the co-occurring temporal adverb do not match.
(1) a. If his son had been born TOMORROWF, John would have been even more pleased.
b. If we had gone out for a walk TOMORROWF, we would have had a good time.
Suppose that John’s son was actually born yesterday, which delighted John enormously.
However, he had secretly hoped that his son would be born two days later (tomorrow), which
happens to be his own birthday. In other words, John was hoping to share the same birthday with
his son, but his wish did not come true. (1a) is used in this situation to indicate that if we assume
counterfactually that his son is born tomorrow, then we can conclude that John is pleased even
more. Morpho-syntactically, an important characteristic of the examples like (1a, b) is that the
2
morphological form of the conditional points toward the past whereas the co-occurring temporal
adverbial indicates a future time.
It is important to see that the evaluation of this counterfactual conditional presupposes
that one and the same person can be born only once. Given this assumption it is easy to see that
two times are competing, as it were, for the time of John’s son’s birth. In other words, the
fictitious situation in question is one in which John’s son is born tomorrow and NOT yesterday.
This example was chosen to highlight the characteristic of this type of counterfactual
(mismatched past counterfactuals, to use the term adopted by Ippolito (2002, 2003)), but I do not
mean to say that an unrepeatable event (like the birth of a child) is needed in a mismatched past
counterfactual. For example, (1b) involves an event that can be repeated at least in principle but
still receives a special interpretation thanks to the temporal adverbial tomorrow. The point here is
that although it is pragmatically possible for us to go out for a walk tomorrow as well as
yesterday, this possibility is excluded in evaluating (1b). One must consider a counterfactual
situation in which we go out tomorrow rather than yesterday. On this assumption, we conclude
that we will have a good time tomorrow. One possible scenario is this: due to scheduling
constraints we have to choose between yesterday and tomorrow for going out for a walk. We
decided to do so yesterday and got rained on. Now the weather forecast says that tomorrow is
guaranteed to be a sunny day.
In both (1a) and (1b), the temporal adverb tomorrow is said to be “focused” in that it has
intonational prominence. This is indicated by a subscripted F in (1). This type of example is not
discussed explicitly in previous proposals such as Stalnaker (1968), Lewis (1973) and Kratzer
(1981, 1989). Rooth (1985) discusses some examples of counterfactuals with a focused
constituent in the antecedent and proposes an account on the basis of Kratzer’s (1981) proposal,
3
but his examples do not involve focused temporal adverbs. This paper presents a proposal that
extends and modifies the Kratzer-Rooth account to accommodate data such as (1a–b).
In “normal” counterfactuals, the same morphological form is used to talk about
counterfactual situations in the past as shown in (2).
(2) If John had been rich, he would have bought a big mansion.
The time of the fictitious situation of John’s being rich is located in the past, and the conclusion
is claimed to follow from this supposition. The situation in (1a) is different in that the fictitious
situation that is entertained is one located in the future. This gives us an interesting puzzle. Why
is it that the grammatical construction that normally specializes in past counterfactual situations
can be used to talk about future counterfactual situations?
Kratzer (1981) expresses her intuitions about counterfactual conditionals as in (3).
(3) The truth of counterfactuals depends on everything which is the case in the world under
consideration: in assessing them, we have to consider all the possibilities of adding as
many facts to the antecedent as consistency permits. If the consequent follows from every
such possibility, then (and only then), the whole counterfactual is true.
English conditionals of the form (4a) normally introduce counterfactual situations located in the
past as in (4b) and are referred to as past counterfactuals.
(4) a. If ... had p.p. ... , ... would have p.p. ...
4
(where p.p. indicates the past participial form of a verb)
b. If John had been rich, he would have been happy.
Present counterfactuals are schematized in (5a) and are exemplified by (5b).
(5) a. If ... V-ed. ... , ... would V. ...
(where V-ed indicates the past tense form of a (stative) verb, and V indicates the
infinitival form of a (stative) verb)
b. If John were rich, he would be happy.
Iatridou (2000) claims that the past tense as used in (4b) and (5b) is “fake” in that it does not
convey the meaning of anteriority. Instead, it has the meaning of being evaluated at a world
different from the actual world. Assuming that past tense makes the same semantic contribution
in (4a) and (5a), we can assume that the past is responsible for conveying a counterfactual
implicature/presupposition. This in turn means that the perfect is used in past counterfactuals to
indicate anteriority.
Iatridou’s account of tense morphemes in counterfactuals is consistent with what is
implicitly assumed in the previous proposals about them. For example, Lewis (1973) posits the
counterfactual conditional connective √∞, which produces formulas such as (6a). This
translates the English counterfactual (6b) and reads officially as ‘If it were the case that John is
rich, then it would the case that he is happy.’
(6) a. John is rich √∞ John is happy
5
b. If John were rich, he would be happy.
This analysis indicates that the past tense in the antecedent and the auxiliary would in the
consequent are used for indicating counterfactuality (a modal concept) and not anteriority (a
temporal concept). The present tense in (6a) or its paraphrase indicates that the temporal location
of whatever the antecedent describes is the utterance time. By extending this general idea, Lewis
informally presents the case of past counterfactuals in the following way.
(7) a. John was rich √∞ John was happy
b. If John had been rich, he would have been happy.
Lewis’s “official English reading” of (7a) is ‘If it were the case that John were rich, it would be
the case that John was happy’.
Given the above informal discussion, it seems reasonable to posit the truth conditions (8)
for (4a) combining the ideas of Lewis (1973) and Iatridou (2000).
(8) ªIf A had V1-ed, B would have V2-edºwc (tc) is defined only if there is a contextually
salient time t < tc. If defined, ªIf A had V1-ed, B would have V2-edºwc (tc) = 1 iff all
worlds w closest to the actual world wc in terms of the similarity hierarchy among those
in which ªA V1(tenseless)ºw (t) = 1, where t is the contextually salient time in wc earlier
than tc, it follows that ªB V2(tenseless) ºw (t) = 1.
6
Note: tc and wc indicate the utterance time (the context time) and the actual world (the
context world), respectively.
(8) accounts for (4b). However, (8) has problems with (1a-b) assuming that the temporal
argument indicates the time of the “situation” or “event” described by each clause. For example,
in (1a, b), the contextually specified past time must be contained within the interval specified by
the adverbial tomorrow, which is a contradiction.
The first problem with (8) is its requirement that the counterfactual situation be located in
the past. (1a–b) violate this requirement because the adverbial tomorrow clearly denotes a future
time. It is not that the proposal given in (8) is completely off the mark, however. Intuitively, (1a–
b) are analogous to other past counterfactuals such as (9), in which a name is focused, in that
they make reference to some relevant past situations.
(9) If BILLF had gone for a walk (yesterday), he would have enjoyed himself.
A counterfactual such as (9) in which a non-temporal expression is focused leads us to believe
that the perfect indicates that what the antecedent describes is located in the past. But given what
we saw in mismatched past counterfactuals, we must arrive at a slightly different conclusion
because it is not necessarily the case that the (possibly) hypothetical situation described by the
antecedent is located in the past. The first hypothesis that we can entertain is that the situation
that is contrasted with the hypothetical situation described by the antecedent is located in the
past. For example, in (1a) the time at which John’s son was actually born is in the past, and in (9)
the time at which someone other than Bill actually went out for a walk is in the past. This
7
account is good for mismatched counterexamples like (1a–b) and also for a “regular
counterfactual” like (9). We shall see below that this account is not general enough. I eventually
hypothesize that the “pastness” associated with (the antecedent of) past counterfactuals is the
position of the contextually salient time. In other words, using a past counterfactual always
indicates that the current “reference time” (Reichenbach 1947) or “topic time” (Klein 1994) is
located in the past. This point will be elaborated below.
Another possible problem with (8) is that it might not make the right predictions as to
which worlds are closest to the actual one in terms of a relevant similarity hierarchy. For
example, it is arguable that in (1b) the worlds in which we go out for a walk tomorrow as well as
yesterday are more similar to the actual world than those in which we go out for a walk
tomorrow but not yesterday. However, the empirical data tell us otherwise. Intuitively, what we
need for (1b) is to imagine a world in which we go out for a walk tomorrow instead of yesterday.
This needs to be explained. I shall eventually explain this in terms of presupposition.
Another fact worth noting here is that the type of interpretation that (1a–b) receive is
truly counterfactual in that (10a–b) would not be appropriate in the contexts in question.
(10) a. If his son was born TOMORROWF, John would be pleased.
b. If we went out for a walk TOMORROWF, we would have a good time.
(10a–b) exemplify what Iatridou (2000) refers to as future less vivid conditionals. They would be
appropriate in circumstances where John’s giving flowers to Mary tomorrow (or our going out
for a walk tomorrow) is unlikely but is not completely ruled out. This type of supposition is very
different from what is found in counterfactual conditionals. More importantly, the proposition
8
described by the antecedent is not contrasted with any proposition that was true in the actual
world. Thus, (10a–b) and (1a–b) cannot be used interchangeably. This is puzzling from the
viewpoint of Iatridou’s hypothesis about “fake past tenses”. In contexts in which (10a, b) are
natural and acceptable, the past tense must be “fake” just as in present and past counterfactuals
because the antecedent is used to talk about a future situation. But there is an important
difference between present and past counterfactuals on the one hand and future less vivids. Put
simply, future less vivids are not counterfactuals in that the speaker believes that whatever the
antecedent describes could take place in the future (though not likely). Although this is an
accurate statement of English facts, it is not clear why “fake past” cannot be used for true
counterfactuals about the future.
2. An Improved Version of the Kratzer-Rooth analysis of Mismatched Past
Counterfactuals (cf. Ogihara 2000)
This section first discusses Kratzer’s (1981) theory of counterfactuals. Then we will see how
Rooth (1985) combines his theory of focus with Kratzer’s theory of counterfactuals to account
for the semantic effects of focus in counterfactuals.
In formalizing her theory, Kratzer (1981) posits a function f that selects for any world w the set
of all propositions that are “the case” in w. We can think of this as an ordering source that orders
the set of possible worlds. Kratzer’s account goes as follows: Aw(p) for any proposition p and
world w is defined as the set of all consistent subsets of f(w)∪{p} which contain p. We can then
say that a counterfactual conditional of the form “if p, q” is true in w iff the truth of q follows
from every maximal set in Aw (p) (assuming that maximal sets of propositions exist in Aw (p)).
9
For practical purposes, the question is which propositions are in every maximal set, and which
propositions are not. Kratzer attempts to “kick out” some propositions by adopting the idea that
some propositions lump together. For example, assume that Hans and Babette spent the evening
together by going to a restaurant called “Dutchman’s Delight.” Given this fact, suppose
counterfactually that Babette had gone to a bistro called “Frenchman’s Horror” instead. We seem
to be justified in concluding from this supposition that Hans would have gone there, too. This
follows from Kratzer’s proposal if we assume that f(wc) (where wc indicates the actual world)
includes the proposition “Babette and Hans spent the evening together.” If this proposition is an
element of f(wc), then this amounts to lumping together the propositions “Hans went to
Dutchman’s Delight” and “Babette went to Dutchman’s Delight.” That is, if one of these
propositions was false, then the other would have to be false as well. Given this supposition, we
can conclude that (11) is true in the situation under consideration as desired.
(11) If Babette had gone to Frenchman’s Horror, Hans would have gone there, too.
Note that (11) contains a focused constituent, namely Frenchman’s Horror. The fact that this
constituent is focused is clearly related to the fact that Frenchman’s Horror is contrasted with
Dutchman’s Delight. In general, when a counterfactual conditional contains a focused
constituent in its antecedent, we can predict which proposition should be gotten rid of by paying
attention to the focus-related information available in the conditional. Let us consider (12a–b),
which differ from each other only with respect to the focused constituents.
(12) a. If John had married SUSANF, he would have been happy.
10
b. If John had MARRIEDF Susan, he would have been happy.
(12a) strongly implicates that John actually married someone other than Susan. We then go
through the following reasoning: Suppose counterfactually that John had married Susan. Then
can we conclude from this assumption that he would have been happy? On the other hand, (12b)
is felicitous only if John actually had some non-matrimonial personal relationship with Susan.
We then reason as follows: Suppose counterfactually that John had married Susan. Can we
conclude that he would have been happy then? Kratzer’s proposal appeals to the lumping of two
relevant propositions in each case to account for its interpretation. But it is not clear where this
assumption comes from. We will get to this question below.
The temporal examples in (1a–b) (repeated here as (13a–b)) have exactly the same
characteristic as non-temporal examples except that the mutually exclusive nature of the two
propositions in temporal examples is prima facie more surprising than non-temporal ones.
(13) a. If his son had been born TOMORROWF, John would have been even more pleased.
b. If we had gone out for a walk TOMORROWF, we would have had a good time.
In Kratzer’s proposal, we would need to posit a more general fact and let it be an element of
f(wc). For instance, in (13a), it might be “A person can be born only once” which is guaranteed
by the pragmatics in this case. Kratzer’s account succeeds in accounting for why some
propositions must be kicked out. However, assuming that we adopt Kratzer’s proposal, we want
to explain how focus facts are related to the selection of the function f since the examples
discussed in the paper are all focus-sensitive. Kratzer’s proposal has been criticized by
11
Kanazawa, Kaufmann and Peters (2005), and Kratzer (2005) responds to this criticism. I believe
that our discussion can proceed without touching upon this recent debate.
Rooth (1985) provides an account of focus effects in counterfactual conditionals on the
basis of Kratzer’s proposal. He examines Dretske’s (1972) examples in (14) and concludes that
the p-set (i.e., the set of alternative propositions) is used in each case to produce an existentially
quantified proposition that is chosen by the partition function f for the actual world wc. The
situation being entertained is one in which Clyde marries Bertha because being married at the
age of thirty is the condition for being eligible for an inheritance.
(14) a. If Clyde hadn’t MARRIEDF Bertha, he would not have been eligible for the
inheritance.
b. If Clyde hadn’t married BERTHAF, he would not have been eligible for the
inheritance.
For each example, we can obtain an existentially quantified proposition with a variable in the
position of the focused expression after removing negation: (15a) for (14a), and (15b) for (14b).
(15) a. There is a relation R such that <Clyde, Bertha>∈R, where R is one of the contextually
salient relations. (Getting married is one of them.)
b. There is an individual x such that Clyde married x, where x is one of the contextually
salient individuals. (Bertha is one of them.)
12
Rooth claims that each proposition thus obtained is an element of f(wc) for the relevant example.
(We will address the question of how to obtain this presumption below.) This means that the
consequent of (14a) is evaluated with respect to a set of worlds in which Clyde dates but is not
married to Bertha (for example). In this situation, it is correct to conclude that the counterfactual
conditional is true. By contrast, (14b) is intuitively false because f(wc) contains the proposition
(15b). This means that the consequent in (14b) follows from the assumption that Clyde married
someone other than Bertha. We can conclude that the counterfactual is false since as long as
Clyde is married, he should be eligible for the inheritance regardless of who his spouse is.
Although it seems artificial to remove negation before calculating the relevant p-set, this
proposal makes the right predictions for the examples (15a–b).
We are now in a position to check whether the Kratzer-Rooth account can be extended to
mismatched past counterfactual conditionals like (13a–b). I believe that a solution to this
problem is suggested by the oft voiced but rarely formalized claim that the contribution of past
tense (in normal declarative sentences) is presuppositional in that it indicates a contextually
specified past time interval that has already been introduced by the time the sentence is uttered.
In a simple declarative sentence, the proposition it describes (i.e., the sentence less the tense
morpheme) is asserted to obtain at this interval. I contend that the same is true of conditionals,
except that the past interval indicated by the perfect (which we could assume has the role of a
past tense morpheme) is not the time at which the proposition described by the antecedent of the
conditional is true. Rather it simply signals the reference time (or topic time). This statement
requires a lot more explanation, which will be given below.
I contend that focus is responsible for yielding a proposition that is contrasted with the
one that is actually given in the antecedent. To be more specific, I contend that the default
13
semantic effect of focused constituents in the antecedent of a counterfactual conditional is
obtained through the adverb instead. It is possible to use a phrase of the form instead of DP
overtly to introduce the contrasted counterpart as in (16a, b), which are based upon (1a, b).
(16) a. If his son had been born TOMORROWF instead, John would have been even more
pleased.
b. If JOHNF had gone for a walk instead, he would have had a good time.
The idea I pursue in this paper is that the semantics of instead interacts with focus even when it
is not there overtly. I assume with Rooth (1992) that focusing causes a focus operator and a
variable to be introduced as a sister node to an expression that contains a focused constituent in
the syntactic representation. Then we can impose a constraint upon the relation between the
ordinary denotation of the focused phrase and its associated variable C.
(17) Where φ is a syntactic phrase and C is a syntactically covert semantic variable, φ ~ C
introduces the presupposition that C is a subset of ªφºf (the focus semantic value of φ)
containing ªφºo (the ordinary semantic value of φ) and at least one other element. (Rooth
1996: 279)
In the case of (1a), we can assume that its antecedent is syntactically represented as in (18). Let
us assume that the tense is ignored for the purpose of calculating the ordinary and focus semantic
values of the antecedent of a conditional. The tense (i.e., the perfect) constrains C, as we shall
see below.
14
(18) S
tu t u S instead (C) tu t u S ~ C wp w p His son is born TOMORROWF
Semantically, the entire clause including the operator instead is interpreted in such a way that it
means that his son is born tomorrow and was not born yesterday. This is empirically correct
except that it does not explain where the counterfactuality of mismatched past counterfactual
conditionals comes from. I thus assume with Rooth that the sentence carries a presupposition that
is obtained by introducing a variable in the position of the focused expression and an existential
quantifier that binds this variable. But there is a further twist in that the reference time is
presupposed to be a past time. If we combine these two presuppositions, we get (19).
(19) ª(1a)º is defined only if there is an interval t ∈ Τ such that John’s son is born within t and
tR contains t, where tR (“reference time”) is a past time and T is the set of contextually
salient intervals that contain tomorrow.
(19) is a stipulation and is not fully formalized, but the idea is this: although both the perfect and
the focused expression occur in the antecedent of a mismatched past counterfactual conditional,
the presupposition associated with the antecedent is projected to the matrix clause. This has an
15
interesting and desirable consequence when we turn to examples that do not presuppose that the
antecedent is false. (20) is one such example due to Anderson (1951:37).
(20) If Jones had taken arsenic, he would have shown just exactly those symptoms which he
does in fact show.
Although the word arsenic receives a focal stress, it does not seem to produce a clear semantic
effect normally associated with focus. In other words, it does not seem to say that Jones took
arsenic instead of something else. If the default interpretation of a focused expression in the
antecedent of a conditional were always made available via the adverb instead, we would not
expect this consequence. Another point to be noted here is that when we say that the antecedent
of (20) is true, this can only mean that it is true at the contextually salient past time indicated by
the perfect. For example, it is hard to interpret (21) in such a way that Jones actually took arsenic
at the time indicated by the temporal adverb the following day.
(21) If Jones had taken arsenic THE FOLLOWING DAYF, he would have shown just exactly
those symptoms which he does in fact show.
This shows that when the antecedent describes an eventuality that obtains at a time different
from the past time indicated by the perfect, then the conditional is truly counterfactual in that the
antecedent must describe a contrary-to-fact situation. Our task is to obtain these generalizations
in a principled manner.
My proposal is given in (22), which improves upon what is given in Ogihara (2000).
16
(22) ªIf DP1 PAST PERF3 VP1, DP2 would PERF3 VP2ºo, wc (tc) is defined only if the
contextually salient time tR < tc. and there is x such that ªDP1 VP1ºf, wc(x)(tR). If defined,
ªIf DP1 PAST PERF3 VP1, DP2 would PERF3 VP2ºo, wc (tc) = 1 iff the proposition {w | there
is an interval i such that ªDP2 VP2ºo (i)(w) = 1} follows from every maximal set in
Awc({w | there is an interval i such that ªDP2 VP2ºo (i)(w) = 1}).
Notes: (a) Assume that a focus semantic value of a clause introduces a lambda that binds
the variable in the focus position. For example, ªJohn meets MARYºf = λx . λt. John
meets x at t
(b) Awc(p) = the set of all consistent subsets of f(wc) ∪{p} that contain p, where f(w) is
the set of propositions that are “the case” in w.
(c) x is a variable of the type associated with the focused expressions.
The idea here is that the proposition that is presupposed is one that must be true at the reference
time, and the reference time is presupposed to be a past time. This means that in a mismatched
past counterfactual that contains a future adverbial in the antecedent, a proposition other than the
one described by the antecedent must be the true proposition. On the other hand, when the
antecedent clause describes a situation that could have taken place in the past, then the true
proposition could be the one described by the antecedent. This means that the antecedent is
allowed to be true in the actual world.
Perhaps the presupposition given in (22) is weaker than what we want. If so we can
17
strengthen it by saying that there is exactly one x (among the contextually salient alternatives)
that satisfies the proposition in question. This mirrors the meaning of the expression instead (of
…) more accurately. This modification can be made very easily if desired.
3. Ippolito’s (2002, 2003) Proposal Examined
Ippolito (2002, 2003) criticizes Ogihara’s (2000) proposal. Let me discuss some of her
counterarguments. First, Ippolito claims that (23a) is acceptable even if we assume that Charlie
had never been to Boston and is already dead. This is unexpected under my proposal, according
to Ippolito since there is no proposition that is contrasted with the antecedent proposition that is
true. Second, a regular past counterfactual such as (23b) could be used to talk about a fictitious
situation in the future. This is also unexpected under my proposal, according to Ippolito.
(23) a. If Charlie had gone to Boston THE DAY AFTER TOMORROW, he would have seen
the Red Sox play.
b. If he had gone to MILAN tomorrow, he would have met my sister.
I argue that in general, past counterfactuals that talk about fictitious future situations are usually
not fully acceptable. The native informant I consulted agrees that (23a) and (23b) are not as
natural as typical mismatched past counterfactuals. Moreover, note that (24a) could be
paraphrased as in (24b), but (24c) cannot be paraphrased as in (24d). This shows some important
difference between them.
18
(24) a. If the Mariners had played the Yankees TOMORROW, I would have gone to the
game.
b. If the Mariners were playing the Yankees TOMORROW, I would be going to the
game.
c. If his son had been born TOMORROW (instead of yesterday), John would have been
even more pleased.
d. If his son were born TOMORROW (instead of yesterday), John would have been
(even) more pleased.
Ippolito (2002, 2003) discusses mismatched past counterfactual conditionals and presents her
own analysis of the construction. In so doing, she argues against my proposal presented in
Ogihara (2000). Her proposal has two ingredients:
(25) Ippolito’s (2002, 2003) proposal (summarized by Ogihara)
a. Accessible worlds are determined in part by the tense form of the conditional. More
specifically, a past counterfactual conditional requires that the accessible worlds are
determined with regard to a past time, and a present (or non-past) counterfactual
requires that the accessible worlds be determined in relation to a non-past time.
b. The context set (the speaker’s presupposition) in the actual world at a relevant salient
time must be consistent with the presuppositions of the antecedent of the conditional.
The “relevant salient time” is determined by the tense form of the conditional. A past
19
counterfactual requires this time be a past time; a non-past counterfactual requires
that it be a non-past time.
(26) a. If Charlie had taken his Advanced Italian test tomorrow, he would have passed.
b. ªSºg,c = 1 iff ∀w∈W[w is accessible from wc at g(2) and Charlie takes his Advanced
Italian test tomorrow in w → Charlie passes in w] defined only if g(2) < tc.
c. # If Charlie took his Advanced Italian test tomorrow, he would pass.
d. ªSºg,c = 1 iff ∀w∈W[w is accessible from wc at g(2) and Charlie takes his Advanced
Italian test tomorrow in w → Charlie passes in w] defined only if tc ≤ g(2).
Regarding (25a) Ippolito’s idea of accessibility is not clear. She paraphrases (26b) in the
following way: for all worlds w such w is a possible future of the actual world at a (contextually
salient) past time and such that Charlie takes his Advanced Italian test tomorrow in w, he passes
in w. The intuition is that for Charlie to take his Advanced Italian tomorrow is still possible
before he actually took it at a past time. On the other hand, (26c) requires that at the utterance
time Charlie’s taking the exam tomorrow is still not precluded. This is not the case since we
know that he already took it at a past time. This is very different from the standard way in which
we think about counterfactuals. We have looked at Lewis’s and Kratzer’s proposals both of
which determine accessible worlds not in terms of whether the antecedent proposition is
possible/feasible in the actual world, but on the basis of the assumption that the antecedent
proposition is true (which is normally counterfactual). In my own proposal, we can choose those
20
worlds that do not contain Charlie’s taking the exam at a past time thanks to the meaning of
instead. My proposal has no problem dealing with examples like (26a).
The point that Ippolito makes regarding (25b) is this: when a non-past counterfactual is
used (which means that the tense form of the antecedent of the conditional is the simple past), the
presuppositions of the antecedent must be compatible with the presuppositions of the context
world at the “contextually salient time” (the context set). Ippolito claims that this is why (27a) is
anomalous, whereas (27b) is acceptable.
(27) a. # Charlie is dead. If he came to the party tomorrow, he would meet Sally.
b. Charlie is dead. If he had come to the party tomorrow, he would have met Sally.
Her explanation is that since Charlie is dead now. A presupposition of "he comes to party
tomorrow" is not satisfied tomorrow, which is that "Charlie is alive". Her proposal is
summarized in (28a, b). The idea is that the antecedent situation must be “possible” at the time
indicated by the tense of the conditional.
(28) P = the set of all worlds in which the conjunction of all presuppositions associated with
the antecedent is true
ct = the context set of any time t
a. Felicity condition for non-past conditionals:
P ∩ ctc ≠ {}, where tc is the utterance time
b. Felicity condition for mismatched past counterfactuals:
P ∩ ct ≠ {}, where t < tc
21
How could we then account for the fact that a counterfactual conditional like (29) is acceptable?
(29) Charlie is dead. If he were here now, he would be enjoying the party.
Just as in (27), at the time of the evaluation (in this case the utterance time), the important
presupposition of the antecedent of the conditional is not satisfied: that Charlie is alive.
Nevertheless the sentence is acceptable.
Note that the difference between (27a) and (29) is not caused by the inherent nature of the
present counterfactual form, but it is caused by the peculiarity of English "future less vivid"
conditionals. This is clear when we turn to Japanese. (30a) and (30b) are both in the same tense
form (past tense) but can talk about different times (future and present times, respectively). Note
that (30a) is not a future less vivid conditional; it is a genuine counterfactual. Unlike English
mismatched past counterfactuals, it does not require a past event that is contrasted with
tomorrow’s hypothetical situation.
(30) a. Mosi asita John-ga koko-ni ki(-tei)-ta-ra,
If tomorrow John-NOM here-at come-PROG-RESULT-PAST
Mary-wa yorokon-da daroo(-ni).
Mary-TOP be-pleased-PAST perhaps
‘If John had come here tomorrow, Mary would have been pleased.’
b. Mosi ima John-ga koko-ni ki(-tei)-ta-ra,
If now John-NOM here-at be-PROG/RESULT-PAST
22
Mary-wa yorokon-de i-ta daroo(-ni).
Mary-TOP be-pleased-PROG/RESULT PAST perhaps
‘If John were here now, Mary would be pleased.’
Note: The combination of mosi and -ra roughly corresponds to the meaning of if or
supposing in English.
We also have difficulty analyzing data that involve fictitious situations the likelihood of which
does not seem to change over time.
(31) a. # If the sun had arisen at 6 a.m. tomorrow, I would have enjoyed the sunrise in
Seattle.
b. (#) If the sun arose at 6 a.m. tomorrow, I would enjoy the sunrise in Seattle.
c. If the sun were to rise at 6 a.m. tomorrow, I would enjoy the sunrise in Seattle.
Most native speakers do not like (31a) or (31b) though some accept (31b). Whatever is the right
judgment, these examples can be used as a test for Ippolito’s proposal.
References
Anderson, Alan Ross (1951) ‘A Note on Subjunctive and Counterfactual Conditionals’, Analysis
12, 35–38.
Dretske, Fred (1972) ‘Contrastive Statements’, Philosophical Review 81, 411–37.
23
Iatridou, Sabine (2000) ‘The grammatical ingredients of counterfactuality’, Linguistic Inquiry
31, 231–270.
Ippolito, Michela (2002) The Time of Possibilities, Truth and Felicity of Subjunctive
Conditionals. Ph.D. dissertation, MIT.
Ippolito, Michela (2003) ‘Presuppositions and Implicatures in Counterfactuals’, Natural
Language Semantics 11, 145–186.
Makoto Kanazawa, Stefan Kaufmann, and Stanley Peters. 2005. ‘On the Lumping Semantics of
Counterfactuals’. Journal of Semantics 22, 129-151.
Klein, Wolfgang (1994). Time in Language, Routledge, London.
Kratzer, Angelika (1981) ‘Partition and Revision: The Semantics of Counterfactuals’, Journal of
Philosophical Logic 10, 201-216.
Kratzer, Angelika (1989) ‘An Investigation of the Lumps of Thought’, Linguistics and
Philosophy 12, 607-653.
Kratzer, Angelika (2005) ‘Constraining Premise Sets for Counterfactuals’, Journal of Semantics
22, 153-158
Lewis, David. (1973) Counterfactuals, Harvard University Press, Cambridge, Massachusetts.
Ogihara, Toshiyuki (2000) ‘Counterfactuals, Temporal Adverbs, and Association with Focus’,
SALT 10 Proceedings, CLC Publications, Cornell University, 115-131.
Rooth, Mats (1985) Association with Focus, Ph.D. dissertation, UMass-Amherst.
Rooth, Mats (1996) ‘Focus’, in S. Lappin (ed.), The Handbook of Contemporary Semantic
Theory, Blackwell, Oxford.
Stalnaker, Robert (1968) ‘A theory of Conditionals’, in N. Resher (ed.), Studies in Logical
Theory, Blackwell, Oxford.