What does it mean to be a possibility?

14

take this full picture into account such as a purely semantic accounts will prove insufficient in accounting for modal expressions when they are examined in discourse.

I will be focusing on the English modal auxiliaries may/might and must. However, the framework should be easily applicable to any language which expresses possibility and necessity through modal expressions. This is because the framework within is nothing but a way of making explicit the implications of the generally-accepted assumptions about language in use described above.

I will show that many of the perceived gaps between use of modal sentences and current linguistic theory stem not from an underpowered or inaccurate theory of semantics or pragmatics, but from a failure to fully take into account how the semantics interacts with the CP and the the common ground over the course of a discourse.

The common ground and the cooperative principle, being alike in that they are external to grammar, are often lumped together by linguistic researchers as “pragmatics,” but keeping their roles separate will be essential in finding a precise analysis of modal utterances.

15

(4) that it snowed in Tokyo in mid-July⁵ Stance: unlikely (5) that Bob left already Stance: possibly likely

(6) that it rained in South Africa Stance: neither unlikely nor likely

Given well-known facts such as that snow usually only occurs during cold weather and that mid-July Tokyo is very hot, (4) is certainly unlikely, but it is not impossible unless you happen to know the weather for every day of mid-July and that it didn't snow on any of them. How about (5)? Imagine that you're at a party, and you're looking for Bob.

Since you don't know where Bob is, the proposition may or may not be likely, but nevertheless you don't know that it is true or you would not be looking for him.

Now consider (6). Assume you are currently in some location whose residents are unlikely to have knowledge about the current weather in South Africa. Can you say that you know that it is not raining in South Africa? Can you even say that it is unlikely that it is raining in South Africa? The proposition seems to be possible in the same broad sense that (4) and (5) are, but there is also a decided intuitive difference in stance—in fact, we might say that there we have no stance regarding (6) other than it is possible in DeRose’s sense.

Although our stances toward (4), (5), and (6) are alike in that we don’t know whether the proposition is true or false, they also differ due to differences in what beliefs we have that we take to be related to them. These related beliefs may be specific to whatever the current situation is, such as not knowing where Bob is at the utterance time, or they may be general beliefs, such as knowing that snow doesn't typically fall in warm climates.

Is there any empirical motivation for or against disregarding these differences in stance (i.e., how they relate to what we know) so that all three propositions can be lumped together as "possible"? Language is an imperfect tool and is not supposed to perfectly represent an individual's thoughts. We would need empirical evidence that these stances play a role in discourse to motivate giving them linguistic expression in a theory of language use.

Below, (7)-(9) put the above propositions under the scope of an epistemic possibility modal. Imagine at what kind of scenes a friend might utter these sentences to you.

5 The that indicates that it is an unasserted proposition.

16 (7) It may have snowed in Tokyo in mid-July.

(8) Bob may have left already.

(9) It may have rained in South Africa.

As we saw, (4) was unlikely but not impossible. However, if we accept (7), in some capacity we accept (4) as more likely than it was prior to the utterance. We get the idea that our friend has some reason to predict that (4) is true, (perhaps he read that it snowed on a marginally-trustworthy online social media website), and feel that it would be misleading for him to utter it otherwise. The same might be said of (5). It wasn't unlikely to begin with, but upon hearing (8) we come to believe that there is indeed some reason to predict that it is true. If we already thought it was likely, we might just agree with our friend. In any case, it would be misleading on the part of our friend if he did not have any such reason. Now consider (9). If a friend who also doesn't know anything about South Africa utters (9) to us, we may have trouble interpreting the utterance. We may be at a loss for why he would say such a thing in the first place. Unlike (7) and (8), the utterance would fail to be misleading even if our friend did not have a reason for predicting that (6) is true because we are not able to accept that she could have one in the first place. (9) is unlikely to give us a sense that we have gained new information.

In sum, we considered very likely contexts in which (4)-(6) are all "possible" in DeRose's sense. According to the typical Stalnakerian conception of assertions, asserting a sentence is an act of trying to get Hearer to accept it as true.

In the case of (7)-(8), the may assertions clearly do more than reiterate that the prejacent is “possible.” We also saw some indication that when Speaker makes a possibility assertion he assumes that Hearer does not already consider the prejacent likely. In contrast, (9) does not clearly do more than reiterate that the prejacent is possible. This should be perfectly acceptable if the role of may were restricted to indicating possibility in DeRose’s sense but, in actuality, it is difficult to even imagine what role in discourse Speaker could have intended for it.

May sentences seem to either make a proposition that was already possible “more” possible, or fail to be an appropriate discourse move. Since there are countless propositions which we do not know are false but about which

17

we have no relevant beliefs, it follows that the number of propositions which can be used felicitously under the scope of possibility modals such as may in any given context are far fewer than the number of those which are semantically

"possible" in DeRose's sense. Added to this, possibility assertions also seem to occur when Speaker assumes Hearer does not already consider the prejacent likely, so there are even fewer semantically-possible propositions that can be felicitously used with possibility modals. Thus, if the truth conditions of possibility operators such as may are as DeRose says, we must seek a non-semantic way to explain the empirical role of the three stances we saw in the above examples.

However, it is also possible that possibility operators have different truth conditions. Thus, there are two possible approaches to finding an explanation for these observations.

A) Change the truth conditions of may so that they entail more than just (DeRose) possibility B) Invent some kind of specialized pragmatic device which is conventionally triggered by may

There have been many endeavors to identify what epistemic modal assertions mean semantically and how they attain their meaning in discourse. Some focus on A) (Kratzer, 2012; MacFarlane, 2003; von Fintel & Gillies, 2010;

Willer, 2013) and some focus on B) (DeRose, 1991; Lyons, 1977; Papafragou, 2006; Portner, 2009; Swanson, 2006; von Fintel & Gillies, 2011), but most agree that both are necessary, to varying degrees. This dissertation adopts Kratzerian ordering semantics, and as such has chosen A) and not B). Of course, Kratzerian ordering semantics alone does not provide an obvious explanation for epistemic modality in discourse or, presumably, more researchers would have adopted it.

The object of this dissertation will be to show how Kratzerian (1981, 1991, 2012) truth conditions interact with a Stalnakerian (1973, 1978, 1996, 2002) discourse model under the Gricean (1989) cooperative principle to predict the observations above and a large number of those from the literature without creating extra pragmatic machinery. It will show that the majority of possibility assertions are attempts to add to the common ground that there is reason to believe that the prejacent is part of a usual state of affairs which is compatible with the utterance context.

18

The new facts revealed by this analysis will require a careful examination of Grice’s maxim of quantity, which demands that a "conversational contribution" be as informative as required, and the maxim of quality, which demands that the “contribution” not be false and be sufficiently justified. By quality, no matter how informative a proposition would be, Speaker must believe it is true in order to assert it felicitously. By quantity, not just any way for a proposition to be true is adequate for a felicitous assertion: it must contribute to a goal.

What is the goal? In the Stalnakerian model of assertion, a successful assertion results in adding a proposition to the common ground. In connection with this, (Stalnaker, 1978) proposes a principle for interaction between

"context and content" such that an asserted proposition not be known to be true or false in the common ground.

This is sufficient to assure that the addition to the common ground of any felicitously asserted non-modal proposition results in a change to the common ground, which is taken to be the basic goal of any assertion.

However, for assertions of possibility propositions, Stalnaker's principle does not offer an adequate accounting. Possibility assertions can be true due to a lack of knowledge or due to a possession of knowledge, but typically only the latter will require a change to the common ground. If felicitous addition to the common ground only requires that the asserted proposition not be known to be true or false in the common ground, a possibility proposition which is true due to a lack of knowledge will always be true, but its prejacent will also not be known to be true or false in most cases, and so there will be nothing to add to the common ground.

We will conclude that Stalnaker’s framework alone is not adequately equipped for modal assertions; we will have to apply Grice’s quantity maxim on top of Stalnaker’s framework to reflect the linguistic fact that possibility assertions are usually informative. The resulting semantic-pragmatic interface will yield new predictions for how discourse involving modal assertions develops, which will be shown to hold in actual discourse.

In the rest of this chapter, I will explain the bare essentials for a basic understanding of Kratzerian modality and provide a bird's eye view of this dissertation's take on how Kratzer's truth conditions, Stalnaker's common ground and the Gricean CP fit together. Detailed formal analysis and comparison with alternative approaches will start in Chapter 2.

19