Sinn und Bedeutung 11 Universitat Pompeu Fabra, Barcelona

A Parallel Account of Epistemic Modals and Predicates of Personal Taste* Tamina Stephenson Massachusetts Institute of Technology

1. Outline • Observation: there are parallels between epistemic modals (might, must) and predicates of personal taste [PPTs] (tasty, fun) related to whose knowledge/evidence (for predicates of personal taste) or taste/experience (for PPTs) is relevant. • Basic proposal: Extend Lasersohn’s (2005) analysis of predicates of personal taste to epistemic modals, with some revisions. • Place the revised analysis within a Stalnakerian theory of conversation and the common ground.

2. The Analytical Puzzle 2.1. Epistemic modals Widely assumed since Kratzer (1977): (1a) ≈ (1b) [epistemic interpretation] (1) (a) (b)

It might be raining. In some world compatible with what is known in the actual world, it’s raining.

BUT – known by who?1 Who is the relevant “knower”? • If “what is known” = what is known by the speaker, the meaning is too weak (see, e.g., MacFarlane 2003). • If “what is known” = what is known by anyone at all or what is known by people in general, the meaning is too strong. • If “what is known” = what is known by A (where A is some contextually salient individual), then we cannot explain certain disagreements (to be discussed below). *

A earlier version of this work appeared in MITWPL 51, New Work on Modality. For comments, discussion, and judgments, I would like to thank Pranav Anand, Kai von Fintel, Danny Fox, Valentine Hacquard, Irene Heim, Sarah Hulsey, Sabine Iatridou, Ezra Keshet, Angelika Kratzer, John MacFarlane, Jillian Mills, Robert Stalnaker, Seth Yalcin, various audiences at MIT, and the editors and reviewers of MITWPL 51. 1 If “what is known” is replaced with “the available evidence,” then the question becomes “the evidence available to who?” This distinction will not matter for my purposes. 1

Epistemic Modals and Predicates of Personal Taste

Much of the literature on epistemic modals is dedicated to this very question, and it turns out not to be easy to answer.2 2.2. Predicates of personal taste [PPTs] Lasersohn (2005) discusses items like fun and tasty. (2a) seems to mean something like (2b). (2) (a) (b)

This roller coaster is fun. This roller coaster gives rise to a certain kind of pleasant experience

But – an experience for who? Whose experience is relevant? Lasersohn shows: • If fun = “fun for the speaker” or “fun for someone” [existential], then the meaning is too weak. • I fun = “fun for everyone” or “fun for people in general,” then the meaning is too strong. The question of whose experience is relevant for PPTs is also difficult to answer. 2.3. Unifying the puzzles My suggestion: We have two very parallel questions which are difficult to answer for very similar reasons, so let’s start by assuming that they are one and the same puzzle.

3. Empirical Parallels between Epistemic Modals and PPTs 3 3.1. Attitude reports Basic observation: When epistemic modals and predicates of personal taste are embedded under think, the relevant knowledge / taste / experience seems to be linked to the subject of think. (3)

Sam thinks it might be raining. = it’s compatible with Sam’s knowledge that it’s raining

2

Besides Kratzer and MacFarlane’s work, see, e.g., Moore (1962), Hacking (1967), Stalnaker (1984), DeRose (1991), Egan, Hawthorne, & Weatherson (2004) [henceforth E., H., & W.], and von Fintel & Gillies (2006). 3 Moltmann (ms.) independently makes similar observations about the parallels between predicates of personal taste and epistemic modals, and gives brief suggestions towards a parallel analysis. 2

Tamina Stephenson

(4)

Sam thinks that the roller coaster is fun.4 = the roller coaster is fun for Sam

Quantified subjects: These facts extend to quantified subjects. (5)

[Every boy]i thinks hei must be stupid.5 = for each boy x, it follows from x’s beliefs that x is stupid

(6)

[Every boy]i thinks hisi dinner is tasty.

3.2. Contradictions and disagreements The Puzzle: A speaker may make and assess assertions involving PPTs / epistemic modals from their own perspective (considering only their own knowledge or experience). (7)

Mary: Where’s Bill? Sam: I’m not sure. (a) He might be in his office. Sue: (b) Nuh-uh, he can’t be. (c) He never works on Fridays. [OR] No, he can’t be. He never works on Fridays.6

(7a) OK if it’s compatible with Sam’s knowledge that Bill is in his office. (7b) OK if it is not compatible with Sue’s knowledge that Bill is in his office. (8)

Mary: How’s the cake? Sam: (a) It’s tasty. Sue: (b) Nuh-uh, it isn’t tasty at all! [OR] No it isn’t, it tastes terrible!

(8a) OK if the cake tastes good to Sam. (8b) OK if the cake does not taste good to Sue.

4. Lasersohn’s Analysis of Predicates of Personal Taste 4.1. A new parameter Lasersohn adds a “judge” parameter (= the individual whose taste/experience is relevant) Crucially, the judge is part of the index: 4

Unlike with epistemic modals, this link turns out not to be obligatory. I’ll return to this in Section 5.2. This example is based on ones from Speas (2004), who argues that the modal base is linked to the bound variable but does not link the effect to attitude predicates. Similar examples are also used in von Fintel & Iatridou (2003) to show that quantifiers can bind pronouns across an epistemic modal. 6 I take expressions like No it isn’t and nuh-uh to mark disagreement in English. 5

3

Epistemic Modals and Predicates of Personal Taste

Kaplan (1989): Character

Content

Truth Value

Context

Index

values of indexicals

world/time of evaluation

4.2. Lexical entries7 Let [[α]]u; w,t,j = the denotation of α w.r.t. context u and index (world w, time t, and judge j)8 PPTs: On Lasersohn’s view, we have the following lexical entries for PPTs (9)

[[fun]]u; w,t,j = [λxe . x is fun for j in w at t] [[tasty]]u; w,t,j = [[taste good]]u; w,t,j = [λxe . x tastes good to j in w at t] [[taste terrible]]u; w,t,j = [λxe . x tastes good to j in w at t]9

Other predicates: Normal, non-judge-dependent predicates do not depend on the judge: (10)

[[be-a-doctor]]u; w,t,j = [λxe . x is a doctor in w at t]

Indexicals: Indexicals depend on the context but not the index: (11)

[[I]]u; w,t,j = the speaker of u [[you]]u; w,t,j = the addressee of u [[now]]u; w,t,j = the time of u

For / to (as used in e.g. fun for Sam, tastes good to Sam) shift the judge parameter10: (12)

[[for]]u; w,t,j = [[to]]u; w,t,j = [λye . [λP . P(w)(t)(y) ] ]

7

I am assuming the following two rules of semantic interpretation, in the style of Heim & Kratzer (1998): (i) If α and β are sisters and [[β]]w ∈ Dom([[α]]w), then [[α β]]w = [[α]]w ( [[β]]w ) (ii) If α and β are sisters and [λw’ . [[β]]w’] ∈ Dom([[α]]w), then [[α β]]w = [[α]]w ( [λw’ . [[β]]w’ ] ) Lasersohn presents his analysis as a self-contained fragment, but the version here is equivalent. 8 Since the judge is an individual, Lasersohn’s index is formally equivalent to a centered world; however, it’s not clear that the judge could be identified with the center given his assumptions about perspective (below). On my revised view this will be possible, and I sketch such an account in the Appendix. 9 Lasersohn doesn’t directly address multi-word expressions like taste good/terrible, and neither will I, but of course these ought to be built up compositionally. 10 Lasersohn (2005, Sec. 5) introduces for syncategorematically with the same effect, essentially stipulating that [[P for x]]u; w,t,j = [[P]]u; w,t,x (where P is a property and x is an individual). 4

Tamina Stephenson

giving us this meaning for (e.g.) fun for Sam: (13)

[[fun for Sam]]u; w,t,j = [λxe . x is fun for Sam in w at t]

= type of function that takes a world, time and judge (the INDEX) to give something of type σ. 4.3. Autocentric and exocentric perspectives Lasersohn’s final assumption: Autocentric perspective: Speakers typically make and assess assertions taking themselves to be the judge. Exocentric perspective: BUT they may take someone else’s perspective if it is salient. 4.4. Consequences Attitude reports: (14)

Sam thinks the dip is tasty.

Sam’s perspective is salient since his attitude is being reported, so an exocentric perspective is possible. Disagreements: Assumption: Speakers disagree iff the content of their utterances contradict each other. [This extends the empirical observation that I’m a doctor / No, you’re not! is a possible dialogue but I’m a doctor / # No, I’m not! is not.] (15)

Mary: (a) Sam: (b) Sue: (b’) [OR]

How’s the cake? It’s tasty. Nuh-uh, it isn’t tasty at all! No it isn’t, it tastes terrible!

[= (8)]

Content of (15a) = f: f() = 1 iff the cake tastes good to j in w at t. Content of (15b)/(b’) = f: f() = 1 iff the cake doesn’t taste good to j in w at t. Æ (15b) contradicts (15a). Compare to: (16)

5

Sam: (a) It tastes good to me. Sue: # (b) No it doesn’t, it tastes terrible!

Epistemic Modals and Predicates of Personal Taste

Content of (16a) = f: f() = 1 iff the cake tastes good to Sam in w at t. Content of (16b) = f: f() = 1 iff the cake doesn’t taste good to j in w at. Æ (16b) doesn’t contradict (16a).

5. Extending the Analysis to Epistemic Modals Basic idea: The relevant “knower” is the judge. 5.1. First attempt (Stephenson 2005)11 Make epistemic modals completely parallel to PPTs: (17)

[[might]]u; w,t,j = [λp . there is some world w’ compatible with j’s knowledge in w at t such that p(w’,t,j) = 1] [[must]]u; w,t,j = [λp . every world w’ compatible with j’s knowledge in w at t is such that p(w’,t,j) = 1]

[simplifying the modal semantics somewhat] This can explain the fact that, like PPTs, epistemic modals: • are linked to the subject of think when embedded • give rise to disagreements with each speaker taking their own perspective 5.2. A problem When embedded under think: PPTs: can be linked to subject (but don’t have to) Epistemic modals: must be linked to subject For example: (18)

Mary: Sam:

How’s that new brand of cat food you bought? I think it’s tasty, because the cat has eaten a lot of it. [Kai von Fintel, p.c.]

[The cat’s perspective is salient, and tasty can be linked to the cat] 11

The proposal in 5.1 is essentially that proposed in Stephenson (2005). Egan (to appear) independently gives what is essentially the same account for epistemic modals, as a form of a relativism. The key difference between Egan’s view (or mine) and previous relativist accounts is that the latter relativize the time of knowledge as well as the “knower” (although as time is not Egan’s focus, it is not entirely clear whether he intends this change to be significant). 6

Tamina Stephenson

On the other hand: (19)

[Context: Ann is planning a surprise party for Bill. Unfortunately, Chris has discovered the surprise and told Bill all about it. Now Bill and Chris are having fun watching Ann try to set up the party without being discovered. Currently Ann is walking past Chris’s apartment carrying a large supply of party hats. She sees a bus on which Bill frequently rides home, so she jumps into some nearby bushes to avoid being spotted. Bill, watching from Chris’s window, is quite amused, but Chris is puzzled and asks Bill why Ann is hiding in the bushes. Bill says:] [E., H., & W. (2004)]

(a)

Bill: I might be on that bus.

[E., H., & W. (2004)]

(b) # Bill: I think I might be on that bus.

[Kai von Fintel, p.c.]

[Ann’s perspective is salient], but might can be linked to Ann in (19a) [for some speakers] but not (19b). 5.3. Revised analysis Take judge dependency out of items themselves: Treat PPTs and epistemic modals as two-place predicates: (20)

[[tasty]]u; w,t,j = [[taste good]]u; w,t,j = [λxe . [λye . y tastes good to x in w at t] ] [[taste terrible]]u; w,t,j = [λxe . [λye . y tastes terrible to x in w at t] ] [[fun]]u; w,t,j = [λxe . [λye . y is fun for x in w at t] ] [[might]]u; w,t,j = [λxe . [λp . there is some world w’ compatible with x’s knowledge in w at t such that p(w’)(t)(x) = 1] ] [[must]]u; w,t,j = [λxe . [λp . for all worlds w’ compatible with x’s knowledge in w at t, p(w’)(t)(x) = 1] ]

New source of judge dependency: Posit a new silent pronoun. (21)

[[PROJ]]u; w,t,j = j

So (e.g.) [[tasty PROJ]] for me = [[tasty]] for Lasersohn and [[might PROJ]] for me = [[might]] on direct extension of Lasersohn. fun for x: Now we can use a kinder, gentler semantics for PPs: (22)

[[to]]u; w,t,j = [[for]]u; w,t,j = [λye . y]

A PPT can directly take a PP as an argument.

7

Epistemic Modals and Predicates of Personal Taste

Silent referential arguments: I assume that there are also null referential pronouns that can simply refer to contextually salient individuals.12 Shorthand: proA = silent pronoun referring to individual A The difference between epistemic modals and PPTs: Simply put, PPTs have the option of taking silent referential arguments and epistemic modals do not. There at least 4 ways of implementing this that I can think of: (I)

As a syntactic selection requirement: might / must are required to take PROJ.

(II)

As a semantic presupposition: might / must have to take an individual argument that is identical to the judge (amounts to the same thing):

(23)

[[might]]w,t,j = [λxe : x = j . [λp . there is some world w’ compatible with x’s knowledge in w at t such that p(w’)(t)(x) = 1] ]

(III) Directly encoding the judge dependency: Keep a Lasersohn-type semantics for might / must (while changing the semantics for PPTs): (24)

[[might]]w,t,j = [λp . there is some world w’ compatible with j’s knowledge in w at t such that p(w’)(t)(j) = 1] ]

(IV) As a semantic restriction on the modal base: epistemic modal bases are inherently judge-dependent: [[Repist]]u; w,t,j = the set of (best) worlds compatible with j’s knowledge I assume (I) purely for convenience. Autocentric perspective: Assume that autocentric perspective is obligatory. (Apparent exocentric interpretations now come from silent referential arguments.) Attitude reports: think obligatorily shifts the judge parameter of the embedded clause to the higher subject: (25)

[[think]]u; w,t,j = [λp . [λxe . for all worlds w’ compatible with x’s beliefs in w at t, p(w’,t,x) = 1] ]13

(This option was proposed in an earlier version of Lasersohn’s paper and later rejected.) 12

Once null referential pronouns are posited, the question arises as to whether these pronouns can be bound in the same way that overt pronouns can. I will have to leave an exploration of their behavior in this respect to future work. 13 This meaning for think does not capture the fact that the shifted judge is interpreted de se (Pranav Anand, p.c., Danny Fox, p.c.). This can done on a centered-worlds view (see Appendix). 8

Tamina Stephenson

5.4. Examples – epistemic modals (26) (a)

It might be raining.

(b)

Structure: [might PROJ ] [ it be raining ]

(c)

[[b]]u; w,t,j = [[might]]u; w,t,j ( [[PROJ]]u; w,t,j ) ( [λw” . [λi” . [λj” . [[it be raining]]u; w”,t”,j” ] ] ]) = [λxe . [λp . there is some world w’ compatible with x’s knowledge in w at t such that p(w’)(t)(x) = 1] ] ( j ) ( [λw” . [λi” . [λj” . it’s raining in w” at i”] ] ]) = 1 iff there is some world w’ compatible with j’s knowledge in w at t such that it’s raining in w’ at t which is equivalent to the meaning from the preliminary analysis (Sec. 5.1)

(27) (a)

Sue thinks it might be raining.

(b)

Structure: [ Sue [VP thinks [S [might PROJ ] [ it be raining ] ] ] ]

(c)

[[b]]u; w,t,j = [[thinks]]u; w,t,j ( [λw2 . [λt2 . [λj2 . [[it might be raining]]u; w2,t2,j2 ] ] ] ) ( [[Sue]]u; w,t,j ) = 1 iff for all worlds w’ compatible with Sue’s beliefs in w at t, there is some world w” compatible with Sue’s knowledge in w’ at t such that it’s raining in w” at t. [if Sue has no false beliefs] = 1 iff there is some world w” compatible with Sue’s knowledge in w at t such that it’s raining in w” at t. [= (26c) with “j” is replaced by “Sue”]

5.5. Examples – PPTs Now there are two options for the structure. (28)

This cake is tasty.

Option 1: Takes PROJ. (Ignoring tense and be; “c” = the cake referred to.)

9

Epistemic Modals and Predicates of Personal Taste

(29) (a) (b)

[ This cake ] [ is tasty PROJ ] [[(a)]]u; w,t,j = [[tasty]]u; w,t,j ( [[PROJ]]u; w,t,j ) ( [[this cake]]u; w,t,j ) = [λxe . [λye . y tastes good to x in w at t] ] ( j ) ( c ) = 1 iff c tastes good to j in w at t which is equivalent to the meaning from the preliminary analysis (Sec. 5.1)

Option 2: Takes null referential argument (e.g., referring to Sam). (30) (a) (b)

[ This cake ] [ is tasty proSam ] [[(a)]]u; w,t,j = [[tasty]]u; w,t,j ( [[proSam]]u; w,t,j ) ( [[this cake]]u; w,t,j ) = [λxe . [λye . y tastes good to x in w at t] ] ( Sam ) ( c ) = 1 iff c tastes good to Sam in w at t which is no longer judge-dependent

Overt PPs: exactly the same as option 2 (31) (a)

The roller coaster is fun for Sam. [like (30)] = 1 iff the roller coaster is fun for Sam in w at t

With think: (32)

Sue thinks this cake is tasty.

Option 1: Takes PROJ. (33) (a) (b)

[ Sue [ thinks [ [ this cake ] [ tasty PROJ ] ] ] ] [[(a)]]u; w,t,j = [[thinks]]u; w,t,j ( [λw” . [λt” . [λj” . [[this cake is tasty PROJ ]]u; w”,t”,j” ) ( [[Sue]]u; w,t,j ) = 1 iff for all worlds w’ compatible with Sue’s beliefs in w at t, c tastes good to Sue in w’ at t [Assuming Sue has privileged access to her own experiences of taste] = 1 iff c tastes good to Sue in w at t

Option 2: Takes a null referential pronoun (e.g., referring to Sam).

10

Tamina Stephenson

(34) (a) (b)

[ Sue [ thinks [ [ this cake ] [ tasty proSam ] ] ] ] [[(a)]]u; w,t,j = [[thinks]]u; w,t,j ( [λw” . [λt” . [λj” . [[this cake is tasty proSam ]]u; w”,t”,j” ) ( [[Sue]]u; w,t,j ) = 1 iff for all worlds w’ compatible with Sue’s beliefs in w at t, c tastes good to Sam in w’ at t which simply expresses a factual (non-judge-dependent belief of Sue’s)

[Note: think still shifts the judge parameter, but it has no effect in this case.]

6. Common ground and the norm of assertion Basis: theory of conversation from Stalnaker (1978, 2002), etc. 6.1. Extending the Stalnakerian picture to judge-dependent framework In conversation: • Common ground = set of world-time-judge triples (instead of just worlds or world-time pairs) • Assertion: proposal to change the common ground by removing world-time-judge triples in which the proposition asserted is not true. [Hearers may accept or challenge the proposal.] • The judge of the conversation = the group of conversational participants [so in this sense there is an “actual judge,” but tasty (e.g.) will not end up just meaning “tasty for us” – see below] Æ If everyone knows who they are talking with, the judge is the same in every member of the common ground. And if everyone also knows what time it is, only the world varies. 6.2. Encoding autocentricity Crucial assumption:

11

Epistemic Modals and Predicates of Personal Taste

• Norm of assertion: In order for speaker A to assert a sentence S, A must believe that S is true as judged by A14 [not as judged by the group of conversational participants] i.e., for all consistent with A’s beliefs, S is true in . [equivalently: A thinks that S, where think shifts the judge parameter] • Accepting an assertion: If A’s assertion that S is accepted, then what is added to the common ground is that S is true as judged by the group of conversational participants [This follows from the definition of the common ground and the role of assertion] 6.3. Examples (35)

Mary: Sam: Sue:

How’s the cake? (a) It’s tasty. [It [tasty PROJ] ] (b) No it isn’t, (c) it tastes terrible!

• (35a) = proposal by Sam to restrict the common ground to triples such that the cake tastes good to j in w at t. [Requirement: for all compatible with Sam’s beliefs, the cake tastes good to Sam in w’ at t’.] • (35b) = objection by Sue to this proposal AND counterproposal to restrict the common ground to triples such that the cake does not taste good to j in w at t. [Requirement: for all compatible with Sue’s beliefs, the cake does not taste good to Sue in w’ at t’.] • (35c) = (counter)proposal to restrict the common ground to triples such that the cake tastes terrible to j in w at t. [Requirement: for all compatible with Sue’s beliefs, the cake tastes terrible to Sue in w’ at t’.] Similarly for (36): (36)

Mary: Sam: Sue: [OR]

Where’s Bill? I’m not sure. (a) He might be in his office. (b) Nuh-uh, he can’t be. (c) He never works on Fridays. (b’)No, he can’t be. (c’) He never works on Fridays. [=(7)]

14

The norm of assertion should probably be formulated in terms of not just belief, but rather something like justified belief. It’s not clear how stricter concepts such as knowledge or true belief should be applied to judge-dependent propositions, and I leave that to future research. 12

Tamina Stephenson

7. Further Predictions 7.1. No “mixed” auto- and exocentric interpretations Key difference between the accounts: accounting for “exocentric” interpretations: (37)

I think the new cat food is tasty, because the cat has eaten a lot of it. [< (18)]

• Lasersohn’s explanation: It’s possible to take the cat to be the judge because its perspective is salient. • My explanation: tasty takes a silent argument referring to the cat. I adopted this view in order to account for epistemic modals, but in fact it has an advantage for PPTs as well. (38) (a) (b)

A: The cat food is tasty. B: (#) No, it’s isn’t!

The facts: • If A intends (38a) to mean “the cat food tastes good to the cat” (and B understands that A meant it that way), then (38b) is only felicitous if B intends it to mean “the cat food doesn’t taste good to the cat.” • On the other hand: If A intends (38a) to mean that the cat food tastes good – in the judge-dependent, autocentric way – (and B understands A’s intention) then (38b) is only felicitous if B intends it in the judge-dependent, autocentric way. My view predicts this since tasty prothe cat and tasty PROJ have different contents. Lasersohn’s view does not predict this, since it’s just a matter of whether speakers choose to take an auto- or exocentric perspective. 7.2. An apparent problem for the revised analysis Problem: Look again at (39): (39)

[Context: Ann is setting up a surprise party for Bill, but Bill and his friend Chris know about it and are spying on her. As she walks down the street carrying party supplies, Ann sees a bus that Bill often rides home on, and runs to hide. Chris asks Bill why Ann is hiding. Bill says:]

I might be on that bus.

13

[=E., H., & W. (2004): (16)]

Epistemic Modals and Predicates of Personal Taste

[for some speakers] might can express Ann’s mental state, so it looks like there is either (1) an exocentric perspective or (2) a referential argument. But: on my proposal, an exocentric perspective is not allowed and epistemic modals cannot take a referential argument! Solution: I suggest that (39) involves ellipsis: (40)

[Context: Same as (39).]

(a) (b)

Chris: Why is Ann hiding in the bushes? Bill: Ann is hiding in the bushes because I might be on that bus.

Suppose that because, like think, shifts the judge parameter (possibly on just one meaning):15 [[because]]u; w,t,j OR [[because2]] = [λp . [λP . [λx . the reason that P(w,t,j)(x) = 1 is that p(w,t,x) = 1] ] ]

(41)

Then we get: (42) (a) (b)

(40b) = [Ann [ [is hiding in the bushes] [because [I might be on that bus] ] ] ] [[(40b)]]u; w,t,j = [[because]]u; w,t,j ( [λw2 . [λt2 . [λj2 . [[I might PROJ be on that bus ]]u; w2,t2,j2 ] ] ] ) ( [λw3 . [λt3 . [λj3 . [[hiding in the bushes]]u; w3,t3,j3 ] ] ] ) ( [[Ann]]u; w,t,j ) = 1 iff the reason that Ann is hiding in the bushes in w at t is that there is some world w4 compatible with Ann’s knowledge in w at t such that Bill is on that bus in w at t.

Æ We don’t need might to take referential arguments.16 [Another possibility: because shifts the judge parameter to something like the person whose reasoning or rationale is relevant.] 15

It’s difficult to give independent evidence for this analysis of because since the kinds of examples that seem to support it may also be explained by binding of a null referential pronoun. (See note 12.) For example, (i) has a reading equivalent to “each boy x is smiling because the food tastes good to x,” which could be explained if the argument of tasty is PROJ and because shifts the judge parameter, but could also be explained if the argument is a null referential pronoun bound by each boy. (i) Each boy is smiling because his food is tasty. 16 A remaining question is why it is that not all speakers accept (46) on the intended interpretation in the first place. This could be taken to either be variation in whether the ellipsis in (65) is allowed, or perhaps variation in whether because can shift the judge parameter, but I’ll leave this open. 14

Tamina Stephenson

Note. We don’t expect this kind of ellipsis to be possible in (43) given (44). (43)

[Context: Same as (39).]

I think Ann is hiding in the bushes because I might be on that bus. (44)

Ann just jumped into the bushes, and Bill and Chris can’t agree on why she did that. # Bill thinks she’s hiding because Chris is there, and Chris thinks she’s hiding because Bill is there.

cf. gapping: (45)

… OK Bill thinks she’s hiding because Chris is there, and Chris, because Bill is there.

8. Summary of Proposal 8.1. Semantic proposal • Lasersohn’s “judge” parameter as part of the index • PROJ refers to the judge • Predicates of personal taste take a syntactic argument (either referential or PROJ) • Epistemic modals always have a judge-dependent meaning (one way or another) • think shifts the judge parameter to its subject 8.2. Pragmatic proposal • The common ground consists of triples • The judge of a conversation is the plurality of the participants in the conversation • The norm of assertion is “autocentric”: the speaker only needs to believe that the content of the assertion is true as judged by them 8.3. Main results • Accounts for disagreements with PPTs (like Lasersohn) and with epistemic modals. • Accounts for interpretation of PPTs / epistemic modals in attitude reports (in a different way than Lasersohn, which proves better).

9. Appendix: Centered Worlds If the judge is an individual, an index is formally equivalent to a centered world.

15

Epistemic Modals and Predicates of Personal Taste

Q. Can the judge be treated as the center of a centered world? A. Yes, I think so. Here is a sketch of how that would work: 9.1. Meanings of judge-dependent items The lexical items would essentially be the same except that “j” in the index would represent the world center. (46)

[[fun]]u; w,j = [λx . x is fun for j in w] [[might]]u; w,j = [λp . p is compatible with what is known by j] (disregarding times)

Note: This meaning for might is exactly what is proposed in Egan (to appear), but different from Egan, Hawthorne, & Weatherson. 9.2. Attitude Reports Roughly: (47)

[[think]]u; w,j = [λp . [λx . x self-ascribes the property of being in a (centered) world w’ where p is true] ] (assume p has the type of a centered proposition)

This will have the same effect as shifting the judge parameter, with the added advantage of accounting for the fact that the judge is interpreted de se (Pranav Anand, p.c.). 9.3. Pragmatic Proposal • Norm of assertion: For a speaker S to assert p, S must self-ascribe the property of being in a (centered) world where p is true. • Common Ground: Set of centered worlds with the plurality of conversational participants as the center. [Note: This means that we must be able to shift in a systematic way between being the center of a world and being part of the center of the world. Another possibility is to follow Egan (to appear) and take the common ground to include centered worlds with each of the conversational participants as center. The effect will be the same as far as what gets added to the common ground when an assertion is accepted.] This means that an assertion of a judge-dependent proposition is a “self-locating assertion.” Egan (to appear) gives arguments for why such a thing might be desirable, and proposes essentially the same analysis for epistemic modals as the one given here.

16

Tamina Stephenson

10. References DeRose, Keith (1991). Epistemic Possibilities. The Philosophical Review, Vol. C, No. 4. Egan, Andy (to appear). Epistemic Modals, Relativism, and Assertion. To appear in Philosophical Studies. URL: http://www.sitemaker.umich.edu/egana/files/might.philstud.final.pdf. Egan, Andy, John Hawthorne, and Brian Weatherson (2004). Epistemic Modals in Context. In Gerhard Preyer and Georg Peter (eds.), Contextualism in Philosophy, Oxford: Oxford University Press. von Fintel, Kai, and Anthony S. Gillies (2006). CIA Leaks. Pacific APA, 2006. First Draft, May 16, 2006. URL: http://mit.edu/fintel/www/cia_leaks.pdf von Fintel, Kai, and Sabine Iatridou (2003). Epistemic Containment. Linguistic Inquiry 34: 173– 198. Hacking, Ian (1967). Possibility. Philosophical Review 76(2): 143–168. Heim, Irene, and Angelika Kratzer (1998). Semantics in Generative Grammar. Oxford: Blackwell. Kaplan, David (1989). Demonstratives. In J. Almog, J. Perry, and H. Wettstein (eds.), Themes from Kaplan. Oxford: Oxford University Press. Kratzer, Angelika (1977). What ‘must’ and ‘can’ must and can mean. Linguistics and Philosophy 1: 337–355. Lasersohn, Peter (2005). Context Dependence, Disagreement, and Predicates of Personal Taste. Linguistics and Philosophy 28: 643–686. MacFarlane, John (2003). Epistemic Modalities and Relative Truth. Ms., UC Berkeley, November 2003. Moltmann, Friederike (ms.). Relative Truth and the First Person. Semantics Archive: http://semanticsarchive.net/Archive/mY2NGJhY/. Moore, G.E. (1962). Common Place Book 1919–1953. London: George, Allen, and Unwin. Speas, Peggy (2004). Person (and Mood and Tense) and Indexicality. Paper presented at the Harvard Workshop on Indexicals, Speech Acts, and Logophors, November 20, 2004. Stalnaker, Robert (1978). Assertion. Reprinted in P. Portner & B. H. Partee (eds.), Formal Semantics: The Essential Readings. Oxford: Blackwell, 2002, pp. 174–161. Stalnaker, Robert (2002). Common Ground. Linguistics and Philosophy 25: 701–721. Stephenson, Tamina (2005). Assessor Sensitivity: Epistemic Modals and Predicates of Personal Taste. In J. Gajewski, V. Hacquard, B. Nickel, and S. Yalcin (eds.), New Work on Modality. MIT Working Papers in Linguistics, Vol. 51. Cambridge, Massachusetts: MITWPL.

17