Truth conditions and non-solipsistic modal bases

Epistemic modality in modal logic was originally taken to be relative to "what is known." Its truth conditions were relative to all things known in the world. However, epistemic modals in natural language are uttered by individuals who do not have access to all things known, and as observations of intuitive truth judgments reflect, are not intended to be relative to this body of human knowledge. Thus, we start with the observation that epistemic modal assertions are subjective, or solipsistic. However, sometimes the truth conditions of a modal sentence seem stricter than what would be the case if Speaker's knowledge were the only knowledge they depended on. This started a quest to find the nature of the knowledge that actually is factored into the truth conditions of epistemic modal sentences in natural language. We will use the following example from (DeRose, 1991) to demonstrate.

A man John has been tested for cancer. His wife and his doctor both know that the results will either determine that he may have cancer, in which case further tests will be necessary, or that he definitely does not. The doctor has seen the results and knows that they are negative, but John's wife has not seen them yet.

(27) John might have cancer.

In this case, John's wife can truthfully utter (27) but John's doctor cannot. Thus, we have evidence that the truth of may depends not on all knowledge, but on Speaker's knowledge. However, if we consider the range of possible answers for John's wife, Jane, it seems we cannot be so hasty to make truth dependent on Speaker's knowledge alone. Consider the following example (cf. DeRose, 1991, pp. 584–585; von Fintel & Gillies, 2011, p. 111).

(28) I've heard that John might have cancer. Is it true?

a (Jane) He might have cancer

b (Jane) I don't know whether he might have cancer; only the doctors know. I'll find that out tomorrow when the results of the test are in.

90

If the truth conditions only require the prejacent to be compatible with Jane's knowledge (and her stereotypical ideals), it should never be the case that she does not know if something is humanly possible or not, and so a response like that in (28b) should be false. von Fintel & Gillies (2011) suggest, "Whatever information state this might quantifies over, it doesn't include just Jane[the speaker]'s knowledge. More objective readings--readings in which the modal quantifies over a modal base that goes beyond the speaker's information state--need to be available" (von Fintel & Gillies, 2011, p. 111)⁵⁴.

DeRose suggests that the modal base is composed of the propositions known by a "relevant speech community." If the community includes only Speaker, or perhaps only Speaker and her family, (28a) is true and (28b) is false. If the community includes the doctors, then (28b) is true and (28a) is false but not knowable by Speaker until she talks to the doctor. Hacking (1967, p. 148) presents a seemingly related problem.

A member of a salvage crew searching for a sunken ship makes some calculations based on a ship log the crew recovered and claims,

(29) "The wreck may be in this bay."

In actuality, an investigation of the logs by a less error-prone third party would have revealed that it was not possible that the wreck was in that bay. Given this information, most will judge that (29) is false. This judgment is constant even assuming that neither Speaker nor anyone in the relevant community knew that the wreck was not in the bay (or that it was not a predictable/normal/usual state of events that the wreck be in the bay). Thus, even if we adopt a modal base which is relative to a community of knowers, (29) is predicted to be semantically true. To account for this gap between truth judgment and semantics, Hacking proposed that the truth conditions of possibility are relative not only to variable communities of knowledge, but also to "practicable investigations." In this way, truth

54 von Fintel & Gillies have identified objectivity of an utterance with the public availability of the propositions in the modal base.

This parts from the use in dissertation, which uses "objectivity" to refer to the act of proposing a proposition for the common ground.

91

can still be relative to a limited body of knowledge such as that of a single speaker, but can also be partially dependent on a more restrictive body of knowledge, such as that of a broader community of which Speaker is a member, or a body of knowledge contains facts which are obtainable through practicable investigations.

To summarize, (29) is claimed to be false because Speaker's claim would not hold up under a practicable investigation. Meanwhile, John's wife is claimed to be able to truthfully assert (28a) if she presents it as relative to only her own knowledge and (28b) if she presents it as relative to the doctor's knowledge, or as relative to her knowledge where there is some practical investigation which she has not undergone (such as opening an envelope containing the test results).

This solution masks a difference between (28) and (29). While John's wife would be able to say (28b) as long as there was a medical test that had been conducted to differentiating her knowledge from the doctor's, we would not typically judge (28a) as false even in the case in which she had an unopened envelope with the results in her hand as she spoke, regardless of what the results happened to be. On the other hand, most would judge Speaker in (29) as having said something false. If opening an envelope and investigating ship logs more thoroughly are both

"practicable investigations," then where does this difference in intuition come from?

Another issue that making the modal base relative to communities tries to address is that even though the truth conditions are relative to some body of knowledge at the time of utterance, Speaker will often admit that he is wrong after gaining new information. If the utterance is relative only to Speaker's knowledge at utterance time, it should not become false by knowledge which is incremented at a later time. Consider the following from (MacFarlane, 2003, 2011).

(30) Sally: Joe might be in Boston. (= It might be the case that Joe is in Boston.)

George: He can’t be in Boston. (= It is not the case that it might be the case that Joe is in Boston.) I saw him in the hall five minutes ago.

Sally: Oh, then I guess I was wrong.

92

MacFarlane argues that if the truth conditions of Sally's utterance were relative only to her own knowledge at utterance time, she should not have to retract her assertion even if further knowledge deems it incorrect. Furthermore, George should not be able to disagree with her in the first place—who is he to say what she knows? If we assume that Sally's assertion was intended and interpreted as relative to her and George's pooled knowledge, on the other hand, then the reason George objects and the reason Sally admits she was wrong can be explained. Sally thought she knew George's knowledge, but George points out that she didn't, and if George says his knowledge is different, Sally has no choice but to admit she was wrong.

Whatever its merits, this explanation leaves us to wonder what made Sally feel entitled to make an assertion based on George's knowledge in the first place. If she intends it as relative to George's knowledge, should she not be obliged to say she doesn't know as Jane was in (28b)? MacFarlane also argues against making the truth conditions of an epistemic modal sentence relative to a relevant community offering the following argument:

Suppose Sally says, “Joe might be in Boston,” and George replies, “Oh really? I didn’t know that.” At this point, Jane—who is hiding in the closet—emerges and says, “Joe can’t be in Boston; I just saw him down the hall.” It seems entirely natural for Sally to reply, “Oh, then I guess I was wrong. Thanks, Jane.” It would be bizarre for her to say, “Thanks for telling us, Jane. I guess Joe can’t be in Boston. Nonetheless, I stand by what I said a second ago.” Clearly Sally did not have Jane in mind when she made her claim. So if we’re going to make sense of these retractions, we must suppose that the force of Sally’s claim was something like: what we know—we who are or will be in a position to consider this claim—does not rule out Joe’s being in Boston. (MacFarlane, 2011, p. 10)

Even when a given knower, such as Jane, is clearly not in Speaker's consideration when she makes her assertion, that knower can disagree despite obviously not being included in the community to which the proposition was intended to be relative, and furthermore Speaker will not disregard this disagreement as inconsequential for her assertion. In short, disagreement by a Hearer and a resultant correction by Speaker can occur even when the disagreement comes from an individual who could not feasibly be included in Speaker's intended community. Thus,

93

wide-community-relativity of the truth conditions does not account for all the observations if the community in question is required to be fixed by Speaker. Nonetheless, without a non-solipsistic reading, it seems we cannot account for felicitous assertions that one does not know whether a proposition is possible or not.

I have summarized the problems below. The leftmost items are empirical observations and the items preceded by an arrow are problems brought about by theoretical assumptions. As can be observed, the observation that epistemic modal assertions depend on Speaker's knowledge is the root of all investigation. The other observations only become problems when this point is acknowledged. When I address these points later, I will assume that the first tier observation is correct, and address the second-tier problems. As such, the third and fourth tier problems do not emerge.

Epistemic modal assertions at least sometimes depend on Speaker's knowledge.

Truth is relative to Speaker's known propositions.

It is possible for Speaker to "not know" whether a proposition is possible.

-> Truth conditions are also relative to the knowledge of larger groups and/or other individuals.

-> Why does Speaker sometimes say "I don't know" and sometimes does not?

We sometimes judge a modal assertion as false even though it is true based on a purely subjective reading

->Truth conditions are also relative to practicable investigations.

What is the difference between ship logs and opening an envelope which leads to different truth judgments?

Speaker will admit he was wrong based on an objection based on Hearer's knowledge.

-> The modal base must somehow include Hearer's knowledge.

-> Why does Speaker feel justified to assert based on Hearer's private

94

knowledge even though he cannot know it?

-> Why can an eavesdropper disagree and why does Speaker acknowledge his point?

In this section, we saw how truth conditional semantics has traditionally approached truth for epistemic modality. First it is observed that Speaker can make an epistemic modal utterance that is judged to be wrong. If epistemic modality is based purely on Speaker knowledge, however, then should almost never be possible since Speaker should know his own knowledge. The fact that it seems wrong can be attributed to epistemic modality being relative to a more exclusive modal base which includes Speaker's knowledge and then some. Even if assumptions of such a modal base makes possible an account for truth judgments, though, it is not argued to explain many discourse effects. These effects are explained through pragmatic devices which operate on top of the community-wide modal base assumption. These proposed explanations are the topic of the next section.