«1 Introduction: Counterfactual Antecedent Falsity (1) If John had come to the party, it would have been fun. (1) is a counterfactual conditional. ...»
LOCAL MAXIMIZE PRESUPPOSITION
Zukunftskolleg and Departments of Linguistics and Philosophy, University of Konstanz
1 Introduction: Counterfactual Antecedent Falsity
(1) If John had come to the party, it would have been fun.
(1) is a counterfactual conditional. Utterances of counterfactual conditionals are typically accompanied by the information that their antecedents are false. But what is the source of that information? Two arguments show that this information is neither an entailment nor a presupposition of the counterfactual. First, counterfactual conditionals can be used to argue for the falsity of their antecedents without begging the question, as in (2). This would be impossible if the ﬁrst premise presupposed or entailed the conclusion. Second, counterfactual conditionals may be uttered when their antecedents are known to be true; the utterer of (3) does not suggest that no one has heard Demosthenes.
(2) If the butcher had done it, he would have used a cleaver. But this wasn’t done with a cleaver, so it wasn’t the butcher (Stalnaker, 1975).
(3) If a man had heard Demosthenes, could he have forgotten it? (Bayﬁeld, 1890).
From this data we infer that the information of antecedent falsity is an implicature, and this paper will examine two problems for an account of how that implicature is generated that was offered in Leahy (2011). The implicature is generated when a speaker chooses to use a counterfactual like (4B’) instead of its past indicative competitor (4B). But what exactly is a counterfactual conditional? A general theory is not available. We can say the following.
Counterfactual conditionals are uniquely in competition with “past indicative” conditionals. That is, there are contexts where one may choose to use either the counterfactual or the past indicative, but where there is no competing non-counterfactual subjunctive. Provisionally, a past indicative is an indicative conditional whose antecedent and consequent both refer to events in the past, as in (4B).
∗Iwould like to thank audiences at MOSS 2 and University of Göttingen, Maribel Romero, Jacopo Romoli, and an anonymous referee.
2 Leahy (4) A: Was Mary’s party any good?
B: If John went, it was fun.
B’: If John had gone, it would have been fun.
C: * If John went, it would be fun.
(4B’) and (4C) are both sometimes called ‘subjunctive conditionals’. In this example, (4B’) is an available alternative to (4B), but (4C) is not. What I am calling ‘counterfactual conditionals’ are those which compete with past indicatives. Some people think that counterfactual conditionals are a subclass of subjunctive conditionals, but I will be silent on this question for now. Noncounterfactual subjunctive conditionals like (5) do not always behave the same as counterfactual conditionals with respect to the implicature of antecedent falsity, and so fall outside the domain of this paper.
(5) If you invited/were to invite me to the party, I would happily attend.
As mentioned, this paper discusses two problems for account of the implicature of antecedent falsity offered in Leahy (2011). One problem arises from the parallel problem of counterfactual consequent falsity: utterances of counterfactual conditionals just as well seem to bear the information that their consequents are false. Again, this seems to be neither an entailment nor a presupposition. There are good reasons to prefer a uniﬁed account of counterfactual antecedent and counterfactual consequent falsity. But we will see that the account proposed in (Leahy, 2011) cannot be straightforwardly extended to the problem of counterfactual consequent falsity.
We also examine the problem of counterfactual antecedent falsity in embedded counterfactuals.
Do utterances of embedded counterfactuals still bear the information that their antecedents are false? We will see that there are conditions under which the account in (Leahy, 2011) fails to account for the observational data.
This paper proposes a common solution to both of these problems. Its structure is this: in the next section I describe the mechanism of antipresuppositions. While accounts appear in (Percus, 2006), (Schlenker, 2006), (Chemla, 2008), and (Sauerland, 2008), I will focus on Chemla’s proposal. In section 3 I will present an account of the presupposition of conditionals, and show that the account provided generates the antipresupposition of counterfactual antecedent falsity.
Section 4 introduces some problematic data for this account regarding utterances of counterfactuals embedded under ‘no’ and then discusses an inconclusive remedy to this problem. Section 5 introduces the problem of counterfactual consequent falsity and demonstrates why the account offered in Leahy (2011) cannot be straightforwardly extended to account for both problems.
Section 6 introduces an alternative mechanism, local maximize presupposition, that resolves both issues. The section begins with a discussion of local maximize presupposition as proposed by Singh (2011). It then shows how maximizing presupposition locally is able to generate the implicatures of both antecedent and consequent falsity. Then it shows how maximizing presupposition locally does, despite a surprising challenge, generate the desired antipresuppositions for counterfactuals embedded under ‘no’. Section 7 concludes.
2 Antipresuppositions In this section I will provide an overview of the mechanics of antipresuppositions.
In Artikel und Deﬁnitheit (1991), Heim explains the infelicity of (6) by appeal to a proposed Gricean maxim, “Maximize Presupposition”.
Counterfactual Antipresuppositions (6) # I interviewed a father of the victim.
(7) I interviewed the father of the victim.
For given that everyone knows that everyone has exactly one father, (6) and (7) make the same contribution to every context. The infelicity of (6) is then explained by the injunction to maximize presupposition when the presuppositionally stronger (7) is an available alternative.
But if Maximize Presupposition is a Gricean maxim, we should be able to exploit it to generate implicatures. And that is in fact observed: B’s utterance in (8) imparts the message that B does not have a girlfriend.
(8) A: Why won’t Betty kiss you?
B: She thinks I have a girlfriend.
(9) Alternative: She knows that I have a girlfriend.
For (8b) has been asserted when (9) is a salient, available, presuppositionally stronger alternative. If conditions are right, the audience will draw the implicature that the speaker doesn’t believe that the stronger presupposition is felicitous, and may further conclude that the speaker believes that the stronger presupposition is false. In the remainder of this section I present the formal apparatus of antipresuppositions as developed in Chemla (2008).
We start with scales of presupposition triggers: that is, ordered sets such that the presupposition triggered by later members of the set are strictly logically stronger than the presupposition triggered by earlier members of the set. Chemla offers data that support the existence of the following scales: a, the, each, the, all, both, believe, know, Ø, again, Ø, too. Suppose a sentence S1 is constructed using an early member of a lexical scale. Suppose that an alternative sentence S2 can be constructed by replacing the early member of the lexical scale with a later member, without changing the assertion of S1. Then, given Maximize Presupposition, it will be inferred that S2 is infelicitous because the constraints on its presupposition are not met. (Note that these are sufﬁcient conditions, though perhaps not necessary. Possible weakenings–most importantly, on the requirement that the alternative make the same assertion–are discussed in Percus (2006).) Chemla then argues that a sentence S with presupposition π can be felicitously uttered by a
speaker s only if:
1. s believes that π is true (Bs [π]);
2. s is an authority about π ((Auths [π]));
3. π is not crucial for the current purpose of the conversation.
 is a variant on a condition that is familiar from traditional scalar implicatures (cf. (Horn, 1972), (Gazdar, 1979)). A speaker’s utterance of a weak sentence compared to an alternative is a reason to believe that the speaker does not believe that the extra information bourne by the stronger alternative is true, ceteris paribus. However, if the extra information bourne by the stronger alternative is presuppositional in nature and not already entailed by the common ground, we must also consider the possibility that the speaker would not be accommodated. Thus we need condition . A speaker can not felicitously use a presupposition-bearing alternative in a context that does not entail that presupposition when her audience would be unwilling to accommodate that presupposition.
4 Leahy Condition  will not concern us here. It exists to explain the data presented in (10) ((21) in
(10) Context: There is a disagreement about the number 319; Mary is known to have very good mathematics skills. Someone just said that 319 is a prime number.
a. *No, Mary knows that it’s not.
b. No, it’s not.
c. No, Mary believes that it’s not.
When a proposition p is under discussion, one cannot felicitously presuppose p.1 Since this condition is always satisﬁed in the cases that concern us here, we will henceforth ignore it.
So if a speaker makes an utterance using a relatively weak member of a scale, her audience may infer that she does not think the stronger assertion would be felicitous. As we are ignoring condition , this means that either the speaker does not believe the presupposition, or she does not believe that she would be accommodated: (∼Bs (π) ∨ ∼Bs (Auths (π))). Then the desired
information follows given three assumptions:
4. Authority: the speaker believes she is an authority w.r.t. π: Bs [Auths [π]]
5. Competence: the speaker has a belief about whether π: Bs [π] ∨ Bs [∼π]
6. Reliability: the speaker may be trusted in her beliefs: (Bs [π]) → π For beginning with (∼Bs [π] ∨ ∼Bs [Auths [π]]), the authority assumption eliminates the second disjunct, leaving us with ∼Bs [π]. The competence assumption strengthens this to Bs [∼π], and the reliability assumption converts this to ∼π.
Note that Chemla’s felicity conditions – are introduced as necessary conditions, while the structure of the argument requires that they are sufﬁcient conditions. I cannot address this problem here; I will simply assume that whatever further conditions are required for joint sufﬁciency are always satisﬁed in the cases at hand. A complete analysis will justify this assumption by spelling out the relevant sufﬁcient conditions.
This section concludes with some illustrations. In uttering (8B), the speaker conspicuously fails to utter an alternative with the stronger factivity presupposition, i.e., that the speaker has a girlfriend. From this the audience may infer that either the speaker doesn’t believe that she has a girlfriend or that the speaker doesn’t believe that she would be taken as an authority about whether she has a girlfriend. If the audience assumes that the speaker takes herself to be an authority, he may conclude that she does not believe that she has a girlfriend. The competence assumption–that she has an opinion on the matter–yields the stronger conclusion that the speaker believes that she does not have a girlfriend. If the hearer takes the speaker to be reliable on the matter, he may conclude that she does not have a girlfriend.2 By way of further illustration, we may note that the corresponding implicature need not arise
from the assertion of (11):
(11) Context: Bill needs a quarter.
Sue (looking through her purse): I think I have a quarter in here somewhere.
1 That is, one cannot presuppose p without generating further pragmatic effects, such as implicating that there is no real room for debate on the issue. This might be the case, for example, if (10a) read, “No, even MARY, the class fool, knows it’s not”, which does presuppose the proposition under discussion and generates a rather bellicose implicature that the discussion is unsound (Chemla (2008)), p. 153-154.
2 In the above, as elsewhere, I treat the speaker (she) as feminine, and the hearer (he) as masculine.
Counterfactual Antipresuppositions Here the implicature of the complement’s falsity will not arise if the competence assumption fails: that is, if Sue suspects, but does not fully believe (in the sense required by the competence assumption), that she has a quarter in her purse.
3 Conditional Presupposition and Counterfactual Antecedent Falsity as Antipresupposition
3.1 The Presuppositions of Conditionals In this section I argue that counterfactual antecedent falsity can be generated as an antipresupposition from an independently motivated account of the presuppositions triggered by various competing conditional constructions. In order for the presuppositions of conditionals to generate antipresuppositions, they must be strictly orderable in terms of logical strength, so as to constitute a scale. If antecedent falsity is to arise as antipresupposition from the assertion of a counterfactual conditional, then the presupposition of counterfactuals must be logically weaker than the presupposition of its indicative alternatives.
Indicative conditionals display some of the behaviours traditionally associated with presupposition triggers. In particular, while they do not assert that their antecedents are consistent with the common ground, it seems that the consistency of the antecedent with the common ground is a precondition on the well-formedness of an indicative conditional. Consider the following
(12) A: John didn’t come to the party.
B: # If John went to the party, it was fun.