2.10 Appendix C: A different approach to strict NPIs and a problem with neg-raising
2.10.4 The solution
The solution that I propose is to weaken the presupposition of want.32 Recall that the presup- position of a wants p adopted above is in (201).
a. only defined iff
∃w0 ∈ f(a, w)[p(w0)]∧ ∃w0∈ f(a, w)[¬p(w0)]
b. when defined true iff
∀w0 ∈BESTg(a,w)(f(a, w))[p(w0)]
I propose now that we should weaken it to that in (202), which only requires that if a thinks that p is possible, then a thinks that also¬p is.
a. only defined iff
∃w0 ∈ f(a, w)[p(w0)]→ ∃w0∈ f(a, w)[¬p(w0)]
b. when defined true iff
∀w0 ∈BESTg(a,w)(f(a, w))[p(w0)]
Notice that both occurrences of p are now in a downward entailing environment, hence no intervention is predicted. There remains, however, the issue that a sentence like (203) in the
31Notice that it is irrelevant for the problem whether or not, given an account of neg-raising, we also conclude
(199), schematized in (200), from (197), as the presupposition part remains the same. (199) John wants Mary not to leave until Thursday.
(200) ∃w0∈ f(j, w)[φ(w0)]∧ ∃w0∈ f(j, w)[¬φ(w0)]∧ ∀w0∈BEST
semantics adopted here would be false, instead of a presupposition failure, if Mary believes that p is not possible.
(203) Mary wants that p
As von Fintel (2004) among others discusses, it is unclear that our intuitions about the differ- ence between presuppositions failure and falsity are reliable, so this new prediction might be defensible.
The presuppositions of soft triggers are
obligatory scalar implicatures
InCHAPTER 2, I introduced the distinction between soft and hard presuppositional triggers, a much discussed topic in the presupposition literature.1 The challenge for presupposition theo-
ries is explaining both the differences and similarities between the two classes of triggers. Let us go through them here briefly. First, soft and hard presuppositions appear to pattern alike with respect to the projection behavior when embedded in propositional connectives. To see this, consider the case of win and it-clefts, a paradigmatic example of a soft and hard trigger, respec- tively. A sentence with win like (1a) gives rise to the inference in (1b) and so do its negation in (1c), a conditional with (1a) embedded in the antecedent like (1d) and the questioned version of (1a) in (1e). Analogously, (2a) and also (2c)-(2d) give rise to the inference in (2d).
(1) a. Bill won the marathon.
b. Bill participated in the marathon. c. Bill didn’t win the marathon.
1see Karttunen 1971, Stalnaker 1974, Chierchia and McConnell-Ginet 2000, Simons 2001, Abusch 2002, 2010,
d. If Bill won the marathon, he will celebrate tonight.
e. Will Bill win the marathon?
(2) a. It is Mary who broke that computer.
b. Somebody broke that computer.
c. It isn’t Mary who broke that computer.
d. If it is Mary who broke that computer, she should repair it.
Second, however, soft and hard presuppositions differ with respect to to whether they can be easily suspended. In particular, the test that I used inCHAPTER2 and that I will use throughout this thesis as the diagnostic for softness is “the explicit ignorance test” by Simons (2001). The idea is creating a context in which the speaker is evidently ignorant about the presupposition; those triggers that do not give rise to infelicity in such contexts are soft triggers. The following two examples modeled from Abusch 2010 show that according to this test win and it-clefts are indeed soft and hard triggers respectively.2
2The boundaries of the soft vs. hard distinction are not uncontroversial (see Abbott 2006 and Klinedinst 2010 for
discussion). There are two main controversial cases: the first case is the status of definite descriptions or possessives, which in some cases like (3) appear to behave like soft triggers, while in others like (4) they do not.
(3) I do not know if Jane has a brother, but if that guy she came in with is her brother the party will be fun (4) I don’t know if Jane has a brother, #but if her brother comes tonight the party will be fun
The distinction between (3) and (4) might be traced back to a difference between argumental and predicative posi- tions (cf. Doron 1983; see also von Fintel 2004). In the following, I put aside these controversial cases, and focus on paradigmatic cases like win and stop.
The second case regards some apparent differences among factive predicates. Karttunen (1971) observes that discoverand regret pattern differently in cases like (5a) and (5b) (from Karttunen 1971): the presupposition that I didn’t tell the truth is suspended in (5a) but appears to go through in (5b).
(5) a. If I discover later that I didn’t tell the truth, I will confess it to everyone. b. If I regret later that I didn’t tell the truth, I will confess it to everyone.
Karttunen (1971) proposes to distinguish between two different classes of factives. Stalnaker (1974) argues that the difference between cases like (5a) and (5b) can, instead, be given a pragmatic explanation. I ignore this issue here and treat all factives uniformly as soft triggers. Notice that the present proposal is compatible with a pragmatic explanation `a la Stalnaker (1974), which can provide a source of difference among soft triggers.
(6) I don’t know whether Bill ended up participating in the Marathon yesterday. But if he won, he will celebrate tonight.
(7) I don’t know whether anybody broke that computer.
#But if it is Mary who did it, she should repair it.
In addition to the speaker saying that she is ignorant about the presupposition, we can also create explicit ignorance contexts, when we make it clear that the speaker does or cannot know whether the presupposition is true. The case in (8) uttered in a conversation between two people who are meeting for the first time is an example of this sort (from Geurts 1995 reported in Simons 2001): (8) is felicitous and the presupposition of stop, i.e. that the addressee used to smoke, is clearly suspended.
(8) I noticed that you keep chewing on your pencil. Have you recently stopped smoking?
Finally, soft and hard presuppositions also differ when embedded in quantificational sentences (Charlow 2009, Fox 2012b; see also Chemla 2009b). The observation, due to Charlow (2009), is that a sentence like (9a) with the hard trigger also typically leads to the universal inference in (9b), while the sentence in (10a), with the soft trigger stop, does not lead to the corresponding inference in (10b).
(9) a. Some of these 100 students quit smoking.
b. Each of these 100 students used to smoke.
(10) a. Some of these 100 students also smoke [Marlboro]F
b. Each of these 100 students smoke something other than Marlboro.
In sum, a theory of the soft-hard presuppositions distinction has to account for the difference with respect to suspendability and the similarity and differences with respect to projection that they exhibit.
Another point of debate is that while the defeasibility of soft presuppositions may suggest an analysis that treats them as implicatures, as I discuss below there are also differences between them, suggesting that any proposal that tries to account for one in terms of the other, has to account for their differences too.
In CHAPTER 2, I have defended a scalar implicature-based account of neg-raising infer- ences building on two ideas which have been put forward in the literature in connection to (soft) presuppositions: first, soft triggers can be thought of as non-presuppositional items associated to lexical alternatives (Abusch 2002, 2010; see also Chierchia 2010). Second, the projection behavior of presuppositions can be given an account in terms of a theory based on scalar im- plicatures (Chemla 2009a, Chemla in preparation). In this andCHAPTER4, I extend this scalar account to give a theory of soft presuppositions. I focus, in particular, on five aspects of the proposal. First, in this chapter, I propose an account for why soft presuppositions are similar and different from hard presuppositions, on one hand, and scalar implicatures, on the other, based on the notion of obligatoriness of scalar implicatures (Chierchia 2006, to appear, Spector 2007a, Magri 2010a). Second, I discuss how the proposal here can account for the projection behavior of soft presuppositions from the scope and the restrictor of quantificational sentences, a combination of predictions that is not made by any of the existing alternative theories that I am aware of. Third, I follow Abrus´an’s (2011a) solution to the triggering problem of soft presuppositions, that is the question about where the presupposition of soft triggers come from. In particular, I adopt and adapt her algorithm in order to give a principled explanation about the source of the lexical alternatives of soft triggers. Fourth, inCHAPTER 4, I show that given its syntactic-semantic nature, the present proposal can account for the intervention effects by soft presuppositions in the licensing of negative polarity items (NPIs) (Homer 2010, Chierchia to ap- pear). As I discuss below, this account of intervention constitutes one of the main arguments for a syntactic-semantic approach like the one proposed here, versus purely pragmatic alternatives in the literature. Finally, I show that the present proposal can also account for some puzzling cases of the interaction between soft presuppositions and “regular” scalar implicatures.
This chapter is organized as follows: in section 3.2, I summarize the three main ideas from Abusch 2002, 2010, Chemla 2009a and Abrus´an 2011b,a, together with some historical prede- cessors in the literature. In section 3.3, I outline the implementation of the proposal and its first predictions. In section 3.4, I show how to account for the differences between soft and hard presuppositions, on one hand, and soft presuppositions and scalar implicatures, on the other. In section 3.5, I discuss the predictions for quantificational sentences, and in section 3.6, I sum- marize and conclude the chapter. Finally, in Appendix A, I discuss the connection between the account of neg-raising inferences outlined in CHAPTER 2 and the one of soft presuppositions presented in this chapter.