What ‘Must’ Adds∗
Matthew Mandelkern
October 11, 2016
Note to Reader: This paper constitutes the second chapter of my dissertation. §4 builds on (and briefly
summarizes) the account of epistemic modals given in the first chapter of the dissertation; otherwise the
work here stands alone. My main points are made in §2-4 (pp. 1-26).
Abstract
Consider the contrast between (1) and (2):
(1)
It must be raining out.
(2)
It is raining out.
There is a difference between the conditions in which one can felicitously use a ‘must’-claim like
(1), versus those in which one can use the corresponding claim without the ‘must’. But it is difficult
to pin down just what this difference amounts to. And it is even harder to account for this difference,
since assertions of pMust pq and assertions of p alone seem to have the same basic goal: namely,
coming to agreement that p is true. In this paper I take on this subtle and long-standing puzzle,
which gets to the heart of questions about the meaning of epistemic modals, the norms that govern
assertions, and the way we process and organize information. I begin by arguing on the basis of novel
cases and experimental data for the empirical generalization that a ‘must’-claim is felicitous only if
there is a shared argument for the proposition it embeds. I then argue that this generalization, which
I call Support, can explain the more familiar generalization that ‘must’-claims are felicitous only
if the speaker’s evidence for them is in some sense indirect. Then I propose a pragmatic derivation
of Support on the basis of a semantics for epistemic modals which I defend elsewhere. I close by
showing how my account can be extended to explain puzzling cases in which a non-modal claim is
unacceptable but a ‘must’-claim is felicitous.
1
Introduction
Consider the following pair of sentences:
(3)
It must be raining out.
∗ I am grateful to audiences at MIT, the 2015 University of Chicago Workshop on Modality and Subjectivity, the New York Philosophy of
Language Workshop, Arché, and Hampshire College; to reviewers for Sinn und Bedeutung 21; and to Dan Baras, Justin Bledin, David Boylan,
Agnes Callard, Nilanjan Das, Brendan de Kenessey, Janice Dowell, Daniel Drucker, Kai von Fintel, Vera Flocke, Irene Heim, Matthias Jenny,
Justin Khoo, Angelika Kratzer, Daniel Lassiter, Rose Lenehan, Sarah Murray, Dilip Ninan, Jacopo Romoli, Bernhard Salow, Ginger Schultheis,
Brett Sherman, Alex Silk, Daniel Skibra, Robert Stalnaker, Matthew Stone, Eric Swanson, Roger White, and Stephen Yablo for very helpful
comments and discussion; to Joshua Knobe and Jonathan Phillips for valuable assistance in experimental design and analysis; and to the MIT
Department of Linguistics and Philosophy for financial support.
1
(4)
It is raining out.
Intuitively, an assertion of (3) and an assertion of (4) have the same basic aim: they are both proposals
to accept that it is raining out. Once an assertion of (3) has been accepted, interlocutors are disposed to
accept the content of (4): that it is raining out. Thus (3) seems to be as strong as (4). But it does not seem
to be stronger than (3), since it is very strange to assert (3) after (4) is already accepted, as witnessed by
the oddness of (5):1
(5) ??It’s raining; and moreover, it must be raining.
This suggests that assertions of (3) and (4) carry the same basic information. Yet the conditions
under which they can be felicitously asserted differ in subtle ways. Suppose that Jane is in a windowless
room, and sees her colleagues come in with wet umbrellas. Then she can assert either (3) or (4). But now
suppose that Jane is looking out a window at the rain. She can still assert (4), but an assertion of (3) −
‘It must be raining out’ − would be decidedly odd.
Generally there exists a systematic difference between the conditions in which one can felicitously
assert a ‘must’-claim with complement ϕ, versus the conditions in which one can felicitously assert ϕ
alone.2 The goal of this paper is to account for this difference − both what it amounts to, and what
explains it.
This puzzle, known as ‘Karttunen’s Problem’,3 gets to the heart of a number of broad foundational
questions. One is a question about the meaning of epistemic modals: how to give an account of the
meaning of epistemic modals which captures the sense in which assertions of pMust ϕq and ϕ have the
same basic goal, while predicting the difference in the conditions in which each can be felicitously used.
Another is a question about the form of the norms that govern assertions − which, in turn, raises the
question of how to give a theory of communication and mind which predicts and explains these norms.
I will argue that these norms − in particular, norms that forbid redundant assertions − have a different
form than has commonly been assumed; once we get clear on their structure, they play a central role in
addressing Karttunen’s Problem. A final question is about the strength of commitment which different
assertions are felt to warrant: in particular, why a speaker of a ‘must’-claim is sometimes felt to make a
weaker commitment to its complement than she would if she asserted its complement on its own.
The main argument of the paper comes in three parts. In §2, I get clear on the data: what exactly the
difference in felicity conditions between sentences like (3) and (4) amounts to. The main claim in the
literature, which I call Indirectness, is that a ‘must’-claim is felicitous only if the speaker’s evidence for
its prejacent is indirect, whereas its bare prejacent can be asserted whether the speaker’s evidence is direct
1 ‘??’
is used to mark felt infelicity, without suggesting anything about the source of the infelicity.
‘must’-claim, or epistemic necessity claim, is a claim (an assertion, or that assertion’s content) containing an unembedded strong epistemic
necessity modal: an expression such as ‘must,’ ‘it is necessary that,’ ‘it has to be that,’ ‘it can’t be that’, etc., read epistemically, unrestricted
by attitude predicates, past tense, or other implicit or explicit restrictions. I will use ‘must’ as an exemplar of such modals. ‘Should,’ ‘ought,’
and their ilk are not strong epistemic necessity modals; I will discuss them below. I will use prejacent to refer to the clause or proposition
that a modal takes as a complement; I assume throughout that ‘must’ takes a prejacent that is not itself a ‘must’-claim. I use ‘ϕ’ as a sentence
variable; ‘JϕK’ denotes the proposition expressed by ϕ (I suppress implicit relativization to contexts for readability). I move freely between
talking about assertions of sentences and assertions of propositions.
3 Following von Fintel and Gillies (2010), who credit Karttunen (1972) with bringing the issue to attention.
2A
2
or indirect. I argue that, while Indirectness is correct, there is another, equally important, generalization
which plays a key role in solving Karttunen’s Problem: namely, that a ‘must’-claim is felicitous only if
the speaker ensures there is a salient argument in support of the claim’s prejacent. I call this constraint
Support. I provide a battery of cases to argue that Support is needed to characterize the difference in
felicity conditions between pMust ϕq and ϕ, as well as experimental results which further confirm this
hypothesis.
In §3, I show that once we have Support clearly in sight, we can derive Indirectness through general
pragmatic reasoning. This reasoning reduces our judgments about the indirectness of ‘must’ to judgments
about when a sequence of assertions is redundant. I argue that this approach is more explanatorily and
empirically adequate than any other extant account of Indirectness.
This discussion involves an excursus on the norm which dictates when an assertion counts as redundant − a norm which, I argue, is at the heart of our judgments about ‘must’-claims. I argue that this norm
differs in important ways from the norm that has been taken for granted in the literature, and sketch a
new framework for thinking about the specific question of why we judge some assertions redundant, and
the broader underlying question of how to model the way we organize and access information.
In §4, I give an account of why Support arises in the first place. I argue against the few extant
accounts, and then propose an account on which Support arises as a Gricean implicature from a semantics
for epistemic modals which I defend in Mandelkern (2016a). The reasoning runs roughly as follows:
pMust ϕq has the same basic update effect as ϕ, but the former raises a question that the latter does not
about the interlocutors’ collective doxastic relation to JϕK. The speaker’s choice to use the more complex
expression thus calls attention to the interlocutors’ collective doxastic relation to JϕK, and thus requires
that some argument for JϕK be made available to the speakers.
This pragmatic approach rightly predicts that assertions of pMust ϕq and ϕ have the same basic
update effect, while explaining why the former, but not the latter, requires that the speaker share an
argument for JϕK, and that her evidence for [[ϕ]] be indirect.
In sum: in §2, I argue for a new empirical generalization, Support. In §3 I show that Support, plus
considerations about redundancy, predicts Indirectness. In §4 I argue for a Gricean derivation of Support.
Finally, in §5-6, I discuss two residual questions: first, how to extend the present account to embeddings; second, how to account for the felt weakness of ‘must’ and the fact that a ‘must’-claim is
sometimes felicitous where the corresponding non-modal claim is not.
2
The Data
I begin by getting clear on the difference in felicity conditions between assertions of pMust ϕq and ϕ.
2.1
Indirectness
The main claim in the literature is that this difference amounts to an indirectness constraint:4
4 Karttunen
(1972), Veltman (1985), Kratzer (1991), von Fintel and Gillies (2010), Kratzer (2012a), Matthewson (2015), Lassiter (2016),
Giannakidou and Mari (2016), Sherman (2016).
3
Indirectness: A claim of pMust ϕq is felicitous only if the speaker’s evidence for JϕK is
indirect; a non-modal claim can be felicitous whether the speaker’s evidence for it is direct
or indirect.
Indirectness is motivated with cases like (6):
(6)
[Watching the rain:]
a. ??It must be raining.
b. It’s raining.
(7)
[Seeing her colleagues enter with wet umbrellas:]
a.
b.
It must be raining.
It’s raining.
(6-a) is distinctly weird as compared with (6-b), (7-a), and (7-b). Indirectness is the most natural generalization to draw from data like these. It has been well-motivated in the literature, and so I will assume
it is correct without further discussion.
More needs to be said about the questions of what counts as ‘indirect’ and what accounts for Indirectness, both questions which I return to below. For now, though, note that I do not assume that the
concept of indirectness which plays a role in Indirectness neatly matches all intuitions about whether evidence is direct or not.5 For instance, reliable testimony is intuitively indirect evidence; but as von Fintel
and Gillies (2010) observe, we must treat it as ‘direct’ when it comes to evaluating Indirectness, in order
to predict the infelicity of assertions like (8-a) in most contexts:
(8)
[Tom tells Susie that it is raining. Susie says to Mark:]
a. ??It must be raining.
b. It’s raining.
Part of what follows will be an attempt to cash out and explain the relevant notion of indirectness.
2.2
Support
Does Indirectness exhaust the difference in felicity between pMust ϕq and ϕ? Most of the literature
on Karttunen’s Problem has indeed focused exclusively on Indirectness. But a different thread in the
literature has pointed to a further contrast in felicity conditions between pMust ϕq and ϕ: in making a
‘must’-claim, the speaker must ensure that an argument for its prejacent is salient to all the interlocutors.
Support: A claim of pMust ϕq is felicitous only if there is an argument for JϕK salient to all
the interlocutors; a claim of non-modal ϕ can be felicitous whether or not there is a salient
argument for JϕK.
5 Or
categories which are encoded as grammatical markers of evidentiality in some languages; see e.g. Willett (1988a), Aikhenvald (2004).
4
Support was first observed in Stone (1994), but it has not received much discussion in the subsequent
literature.6 The data that motivate Support are indeed less clearcut than those that motivate Indirectness.
This is unsurprising: evaluating Support requires evaluating discourses as a whole, rather than single
utterances, and it can be quite difficult to determine, in a given context, whether an argument has been
made salient. Both these facts produce a fair amount of noise in evaluating Support. In the remainder of
this section, I will provide new data to argue that Support is indeed required to account for the difference
in felicity conditions between pMust ϕq and ϕ.
Consider the following case:7
(9)
Patch the rabbit sometimes gets into the cardboard box where her hay is stored. On his way out
the door, Mark hears a snuffling from the box and thinks to himself, ‘Patch must be in the hay
box.’ When he gets to school, Bernhard asks him how Patch is doing.
a.
b.
[Mark:] She’s great. She must have gotten into the hay box this morning.
[Bernhard:] Cute!
Suppose the conversation ends here, and assume that Bernhard doesn’t know anything about Patch’s setup at Mark’s house, or in general anything which might help him figure out why Mark thinks that Patch
was in the box of hay. There is something distinctly odd about this exchange. Intuitively, what Mark has
said needs some more elaboration; either Mark should have proffered reasons to think that Patch was in
the hay box, or Bernhard should have asked him for reasons, perhaps by saying, ‘Why do you say that?’
Here is a more felicitous version of (9):
(10)
a.
b.
[As in (9), except Mark says] She’s great. I heard a snuffling from the box of hay on my
way out − she must have gotten into the box.
[Bernhard:] Cute!
Now suppose the conversation ends here. This exchange has none of the peculiarity of (9).
What does this show? First, note that a non-modal variant of (9) is perfectly fine:
(11)
a.
b.
[As in (9), except Mark says:] She’s great. She got into the box of hay this morning.
[Bernhard:] Cute!
The non-modal variant on (10) is likewise fine. The infelicity of (9) thus seems to be due to the use
of ‘must’. But note that in both (9) and (10), Indirectness is satisfied, and yet (9) seems strange. So
Indirectness cannot explain the contrast between them. By contrast, Support predicts precisely the pattern
we observe: namely, that (9) will be weird − since no argument for the prejacent is made salient − while
(10) and (11) will be perfectly acceptable.
Cases like this thus provide our first piece of evidence for Support. In a moment I will give a variety
of cases which elicit similar intuitions. But first we should say more about what Support amounts to.
6 It
is taken up briefly in Murray (2014), Silk (2014), Swanson (2015), and Lassiter (2016).
will leave off double question marks in this section, since the infelicity in question is attached to the discourse as a whole, and not to any
one sentence.
7I
5
First, what does an argument amount to? I will think of an argument for JϕK in a particular context as a
set of propositions which the speaker is commonly recognized to believe provides reason to believe JϕK
− either by deductively entailing its conclusion; by inductively supporting the conclusion; or by showing
how the conclusion follows from what is already accepted.
Second, what does ‘salience’ amount to? I won’t say much about this, but a few features are worth
noting. First, an argument need not itself be common ground − the set of propositions commonly accepted in the conversation.8 One can felicitously assert an argument conjoined with a ‘must’-claim, even
if the argument has not yet been (and never is) accepted by all the speakers (if Bernhard doesn’t believe
me that I heard a snuffling from the box of hay, this does not render (10) infelicitous). The sense in which
an argument Γ must be salient is rather that it must be common ground that the speaker takes Γ to provide
reason to believe the prejacent of her ‘must’-claim, and that she is proposing to add Γ to the common
ground. I will refer to an argument with this status as ‘salient’ or ‘shared’ or ‘publicly available’.
An important point about salience is that an argument can be salient without being made explicit, as
in (12):
(12)
[Bernhard and Mark are in the bunny’s room, and can both hear snuffling from the box of hay.
Mark:] Patch must be in the hay box.
Here, the premise that merits Mark’s conclusion − that Mark can hear snuffling from the box − is salient,
and the ‘must’-claim is acceptable.
Another noteworthy feature of the notion of salience in question is that the argument in question
need not be salient at the time of the assertion; it can be provided shortly after the assertion.9 Note, for
instance, that (13) is felicitous:
(13)
2.3
a.
b.
c.
[As in (9), but Mark says] Patch must have gotten into the box of hay.
[Bernhard:] Why do you say that?
[Mark:] I heard her snuffling around when I was leaving.
More Cases
We find further confirmation of Support when we turn our attention to a broader range of cases. Consider
e.g. (14), adapted from Murray (2014):
(14)
Sarah works in a windowless building. On her way to a meeting, she sees her coworker Jim
enter the building, carrying a wet umbrella. Sarah concludes from this that it’s raining out.
Sarah enters the meeting. Her colleague Thomas, who didn’t see Jim carrying a wet umbrella,
asks, ‘What’s the weather like?’ Sarah responds:
a.
b.
It must be raining out.
It’s raining out.
8 See
e.g. Stalnaker (1970, 2002, 2014); I follow the formulation in the latter.
matches patterns for other constructions that require something to be made salient, like the resolution of referents for demonstratives,
which can also be made salient after an assertion.
9 This
6
c.
d.
It must be raining out; I just saw Jim come in with a soaking wet umbrella.
It’s raining out; I just saw Jim come in with a soaking wet umbrella.
Thomas replies: ‘Oh, too bad. Ok, let’s talk about the agenda for this meeting.’
As Support predicts, (14-a) is odd, while the other variants are fine.
We find a similar pattern in (15):
(15)
Jane is in her first year of college. She doesn’t have a clear sense of how she is doing in school.
She meets with her professors, who tell her she is doing well; she thus concludes that she is
doing okay. She goes in to meet with her adviser to talk about course registration. Her adviser
doesn’t know about the conversations she’s had with her professors. Her adviser asks: ‘So, how
are you doing in your classes?’ Jane responds as follows:
a.
b.
c.
d.
I must be doing okay!
I’m doing okay!
I must be doing okay: I’ve spoken to all my professors and they told me I’m doing fine.
I’m doing okay: I’ve spoken to all my professors and they told me I’m doing fine.
[Her adviser replies:] ‘Good, I’m happy to hear that. Ok, on to our business for today: let’s discuss
your registration for next term.’
Again, as Support predicts, (15-a) is marked, while the other variants are fine.
Two final examples nicely Support, and in particular show that the argument in question needs to be
shared even if it is not spelled out explicitly, but instead is quietly accommodated. (16) is taken from a
classical music radio show:
(16)
Mozart wrote the Stadler quintet for his friend Anton Stadler, who must have been a marvelous
clarinetist.
The announcer does not give explicit reasons in support of the claim that Anton Stadler was a good
clarinetist, but they are easy to recover from the context (the difficulty of the piece, the fact that Mozart
wrote it for him).
The second illustration: my phone rings; I can say:
(17)
This must be my brother; let me take this.
It is, again, easy for you to recover my reasons for saying this: that I am expecting a call from my
brother.10
10 Support
is also evidenced in written and pictorial forms. Lassiter (2016) discusses a range of examples taken from genealogical discussion
boards, noting that ‘Ancestry.com users frequently provide an explicit specification of the evidence used to arrive at a must conclusion.’
Here’s one example Lassiter discusses, taken from The Plymouth Colony Archive Project.
(18)
Goodman. . .is listed as one of those who received land in 1623 (PCR 12: 4). He is not listed among those who were part of the
cattle division of 1627, so he must have died by then.
7
Comparing ‘must’ with other words that might at first glance seem to work in a similar way can help
bring out the plausibility of Support. Consider (19), adapted from a television spy drama:
(19)
a.
b.
The suspect is fleeing south. We’ve sent agents ahead to Mattapan.
Why Mattapan?
(i) Apparently the Russians have a safe-house there.
(ii) The Russians must have a safe-house there.
If the conversation ends here, then (19-b-ii) is peculiar in a way that (19-b-i) isn’t. ‘Apparently’, like
‘must’, is constrained by a form of Indirectness; but ‘apparently’, unlike ‘must’, is acceptable here
without an argument.11 Support predicts precisely this contrast (assuming that no corollary governs
‘apparently’).
Finally, I note that informal polling suggests that the contrasts reported here are robust across strong
epistemic necessity modals, in English and in other languages.12
2.4
Experimental Data
The examples given in the last section manifest a contrast precisely where Support predicts one for
almost all the speakers I have polled informally. In order to confirm these intuitions, I ran an experiment
asking subjects to judge the felicity of the four different variations on (14) and (15) from the last section.
The variations differ, again, according to whether the claim in question is modal or non-modal, and
according to whether an argument is given or not. Each of 153 Amazon Turk subjects was given all
four variations of one of the scenarios, in randomized order.13 For each variation, subjects were asked to
respond to the statement
(20)
There is something weird about this exchange.
on a scale of 1-7, with 1 representing ‘completely disagree’ and 7 representing ‘completely agree’. Figure
1 shows the rating of weirdness for each possible combination of the two conditions.
Collapsing results from the two scenarios, I found an interaction (F(1,584) = 3.849, p = 0.05) between whether the statement included ‘must’ (as in the first and third continuations of (14) and (15))
and whether there was an argument proffered (as in the third and fourth continuations). This interaction was primarily driven by the difference in perceived weirdness between ‘must’-claims with an
argument (mean=1.926, SD=1.33) and those without one (mean=3.3423, SD=1.89; t(265.43) = 7.49,
p<0.001, d = 0.868). In contrast, while there was still a difference in perceived weirdness between nonmodal statements with an argument (mean=1.8993, SD=1.34) and non-modal statements without one
A variant which omits the first half of the second sentence is felt to be omitting something important. By contrast, a non-modal variant which
omits the first half of the second sentence sounds fine. Likewise, consider a t-shirt with a club logo on the front, and on the back a schematic
picture of people dancing, with the slogan ‘This must be where the party is’. Now imagine a shirt with the same front, but with no picture on
the back, only the slogan by itself. This would be completely puzzling. By contrast, the non-modal version of the shirt − one which simply
says ‘This is where the party is’ − would be (banal, but) unpuzzling.
11 I assume that the indirectness constraint for ‘apparently’ is somehow lexically encoded. Thanks to Justin Bledin for discussion.
12 Informants report the predicted contrast in Bengali, French, German, Hindi, Japanese, Russian, Spanish, Swiss German, and Turkish.
13 On the use of Amazon Turk to elicit judgments in linguistics, see e.g. Gibson et al. (2011).
8
Rating of Weirdness
3
2
Argument
No Argument
1
0
Must
No Must
Figure 1: Mean ratings by condition. Error bars show standard error of the mean.
(mean=2.3892, SD=1.44), the difference was much smaller (t(294.56)=3.0449, p<0.005, d=0.353).
2.5
Discussion
These results provide further evidence for Support: subjects regard epistemic necessity claims without
arguments to be substantially stranger than epistemic necessity claims with arguments, or non-modal
claims with or without arguments.
The results of the experiment require further comment in two respects. First, the data suggest a
more continuous picture than Support suggests: they suggest that ‘must’-claims have a much stronger
requirement than non-modal claims that their prejacent to be supported by shared evidence − but that
there may be mild pressure to share one’s evidence for non-modal claims as well. This graded version
of Support is, I think, compatible with everything I say in what follows, and so could be substituted
throughout. For simplicity, however, I will stick with the binary presentation of Support I gave above.
Second, although subjects judged ‘must’-claims without arguments to be substantially worse than
any of the variants, their infelicity judgments for these claims were roughly at a midpoint. This suggests
that failure to conform with Support leads to some markedness, but not complete unacceptability. This
can be explained by a few factors, two of which I mentioned at the outset. First, again, to bring out the
infelicity of these cases, we must judge entire discourses, rather than single assertions. This makes the
task harder than the corresponding task for Indirectness.
Second, when it comes to a constraint like Support, there are two further sources of noise. The first
9
is accommodation: as we have seen, an argument can become salient to a speaker’s audience through
implicit clues. In cases like this, no argument is given explicitly, which may make them look like counterexamples to Support. But an argument is made salient, which the audience can easily reconstruct.
Second, it is often perfectly clear to us what the speaker is trying to do with a ‘must’ claim, even if she
flouts the Support requirement. If a speaker asserts pMust ϕq, we know that she is trying to update the
common ground with JϕK (more on this below). We will experience her assertion as marked if there is
no salient argument for JϕK, but we still know what she is trying to do with it, and we may simply not
care about the fact that it is marked. In this respect Support stands in stark contrast to a felicity constraint
like the requirement that pronouns have salient anaphors: if that constraint is flouted, we may simply not
know what the speaker is trying to accomplish with her assertion.14 In light of all this, it is not surprising
that violations of Support are judged to be infelicitous at only intermediate rates.
I will ultimately try to make sense of these data by arguing that Support has the status of a manner
implicature. But that is getting ahead of ourselves; for now, I conclude simply that we should adopt
Support as part of our characterization of the difference in felicity conditions between a ‘must’-claim
and its bare prejacent.
3
Explaining Indirectness Via Support
We are now clear on the data: Indirectness and Support are both necessary to characterize the difference
in felicity conditions between a ‘must’-claim and its bare prejacent.15
We now turn to the question of how to explain these data. There are three strategies to consider:
1. Account for Indirectness and Support separately.
2. Account for Support in terms of Indirectness, and give an independent account of Indirectness.
3. Account for Indirectness in terms of Support, and give an independent account of Support.
In this section I will pursue the third strategy, after briefly arguing against the first two strategies.
3.1
Against Strategies 1 and 2
Considerations of theoretical parsimony tell against pursuing the first strategy − providing separate
explanations for Indirectness and Support − unless these separate explanations bottom out in a unified
theory. I do not see an attractive way to pursue this strategy, and I therefore set it aside as a last resort.
What about the second strategy? This strategy is prima facie attractive, since there are a number
of extant attempts to give an independent account of Indirectness; it is natural to try to recruit them to
explain Support.
14 Of
course, if the pronoun is being used in an aside of some kind, then we may not worry about the lack of a salient anaphor. Likewise, ‘must’
can be used in asides, in which case we may not worry about the lack of a salient argument. For instance there is a common usage of ‘must’
in first person reporting along the lines of ‘It must have been ten years ago’, where the speaker is felt to be in part engaged in some kind of
dialogue with herself as she pulls together her recollections.
15 I will not settle the further question of whether they are jointly sufficient, which is interesting but does not matter for present purposes.
10
But there are two significant problems with this approach. First, there does not seem to be any way
to reduce Support to Indirectness. A natural first thought is that we can explain Support in terms of
Indirectness by way of a general pragmatic constraint that requires a speaker to share her evidence for
a claim if that evidence is indirect. But there is no such pragmatic constraint, as we saw in cases above
where non-modal claims were felicitous without shared evidence. A closely related thought is that there
is a general pragmatic constraint which requires a speaker to share her evidence if she explicitly indicates
the source of her evidence.16 But, again, as the example above with ‘apparently’ shows − and as crosslinguistic work on evidentials suggests (see Murray (2014)) − there is no such constraint: again, one can
use ‘apparently’ or evidential marking without sharing what your evidence is.
A natural second thought is that Support reduces to a requirement to assure your interlocutors that
Indirectness is satisfied. But this approach is not plausible, for a few reasons. First, in most of the cases
given above that are felt to be infelicitous without an argument − Patch in her box, Sarah in her windowless office building, the Russian safe-house − there is simply no reason to worry that the speaker’s
evidence might not be indirect. Second, it is not generally true that whenever a formulation is constrained
by a form of Indirectness, the speaker must habitually share her evidence in order to reassure her interlocutors that it satisfies the constraint in question: again, we saw this in (19) with ‘apparently’, which
is governed by an Indirectness constraint, but which doesn’t require a shared argument.17 Finally, from
a more theoretical standpoint, it is hard to see why an Indirectness constraint would ever directly yield
an obligation to share one’s evidence: we are fairly charitable in assuming that speakers are complying
with felicity conditions. For instance, if Indirectness were encoded as a presupposition − on which more
in a moment − then, on a standard approach to presuppositions, it will be required that it be common
ground that the speaker’s evidence for the prejacent is indirect. But in general interlocutors are perfectly
happy to accommodate presuppositions.18 To derive Support from Indirectness in this way, we would
need an explanation of why interlocutors are not in general willing to accommodate an Indirectness presupposition, and instead require that the speaker’s evidence be spelled out so that they can verify that it’s
indirect. Perhaps something along these lines can be spelled out, but I do not see how to do it.
This does not close the door to a derivation of Support from Indirectness; but at present I do not see
a promising way for this to go.
The second problem with the second strategy is that explaining Indirectness itself has turned out
to be quite tricky. Both major extant explanations have well-known drawbacks, which I will briefly
survey here. So it is not clear that, even if we could reduce Support to Indirectness, we would yet have a
satisfactory solution to Karttunen’s Problem.
There are two major extant approaches to explaining Indirectness. The first approach is pragmatic.
On this approach, an assertion of pMust ϕq is pragmatically weaker than an assertion of ϕ, in a sense
I will make precise in a moment. By choosing the weaker pMust ϕq, rather than the stronger ϕ, the
speaker signals that her evidence for JϕK is too weak to merit an assertion of ϕ. Thus, in turn, we can
16 Thanks
to Justin Bledin for discussion.
again, Murray (2014) likewise observes that grammatical evidential markers for indirectness do not give rise to any obligation to share
one’s evidence.
18 See e.g. Lewis (1979), Stalnaker (2002) and many others.
17 And,
11
conclude that the speaker’s evidence for JϕK is indirect, since direct evidence would have merited an
assertion of ϕ.19
This approach seems obviously correct for capturing the correlate to Indirectness for weak epistemic
modals: modals like ‘might’ and ‘probably’, as well as for ‘should’ and ‘ought’ if we classify those as
epistemic modals.20 Assertions involving an unembedded weak epistemic modal are uncontroversially
weaker than the corresponding non-modal assertions, and thus this account is a natural explanation of
the corresponding indirectness inference. Nor, in those cases, does anything like Support arise (as I
will discuss further in §4.2 below), which makes this pragmatic approach a satisfying explanation of
Indirectness for those cases. But it is not clear that this approach extends equally well to the Indirectness
of ‘must’: the assumption that pMust ϕq is pragmatically weak is at odds with many intuitions about the
relative strength of ‘must’-claims, intuitions which seem better captured instead by Pragmatic Strength:
Pragmatic Strength: An assertion of pMust ϕq makes a proposal which is just as strong as
an assertion of ϕ, in the sense that once the common ground is updated with JMust ϕK, it is
updated with JϕK.
Given the standard assumption that the common ground is meant to track collective behavioral disposition in some sense, Pragmatic Strength essentially says that conversants do not typically leave open the
possibility of J¬ϕK in their behavioral dispositions after accepting pMust ϕq. To see its plausibility, note
that (21-b) and (21-c) are decidedly weird responses to (21-a):
(21)
a.
b.
c.
The gardener must be the murderer.
I concur. Moreover, the gardener is the murderer.
I concur. Let’s bring him and the butler in to see if we can pin down which of them is the
murderer.
Pragmatic Strength directly explains the infelicity of (21-b) as a response to (21-a). And the assumption
that ‘must’ is pragmatically strong also explains the oddity of (21-c): once a group of interlocutors
accepts pMust ϕq, they are committed to accepting JϕK in their collective behavioral dispositions. (In
§6 I argue that there is also a sense in which ‘must’ is weak − a sense which involves less speaker
commitment to JϕK when it appears embedded under ‘must’ than when unembedded; my proposal there
captures this intuition in a way that is compatible with Pragmatic Strength.)
I will not argue extensively for Pragmatic Strength here; I believe that data like (21) speak strongly
in its favor, and, though it has been assumed to be false in pragmatic derivations of Indirectness, it
has not as far as I know been directly challenged in the literature. von Fintel and Gillies (2010) give
a battery of arguments in support of a stronger thesis − that ‘must’ is semantically strong − which
in turn entails Pragmatic Strength.21 Lassiter (2016) challenges this stronger thesis, but not Pragmatic
19 This
approach is given in Veltman (1985), Kratzer (1991).
everyone does; see Yalcin (2016b).
21 von Fintel and Gillies (2010) argue that ‘must’ is semantically strong in the sense that pMust ϕq entails that the contextually relevant agents
know ϕ, and thus that ϕ is true. I remain agnostic about this stronger, more controversial hypothesis, which entails, but is not entailed by,
20 Not
12
Strength; indeed, he presents experimental data which show that, in a case he formulates, subjects agree
with pMust ϕq and ϕ at about equal rates − a prediction in line with Pragmatic Strength, but prima
facie at odds with its negation. If Pragmatic Strength is correct, however, then the present pragmatic
derivation of Indirectness does not work.
The main alternative explanation of Indirectness, due to von Fintel and Gillies (2010),22 is to posit
that epistemic modals lexically encode that the speaker’s evidence is indirect.
This approach has three theoretical drawbacks which should give us pause. First, as von Fintel and
Gillies themselves note, it does not look particularly explanatory, given the robustness of Indirectness
for ‘must’-claims, both within and across languages. Second, it turns out to be quite difficult to pin
down the type of evidence ‘must’ selects for within standard taxonomies of evidentials, suggesting that
categorizing ‘must’ as a lexical evidential may not be quite right.23 Third, this approach forces us to
take a disunified approach to modality, since other modals (deontic, circumstantial, etc.) do not carry
evidential marking.24
A more decisive worry about this approach is that it does not make the right predictions about
embedded epistemic modals. On von Fintel and Gillies (2010)’s approach, Indirectness is encoded as
a lexical presupposition of ‘must’ that the relevant direct evidence (what they call the kernel) does not
directly settle the prejacent or its negation: i.e., no single piece of the agent’s direct evidence entails
the prejacent or its negation.25 In order to capture the fact that negated epistemic possibility modals
like ‘cannot’ carry Indirectness to the same extent as ‘must’, von Fintel and Gillies (2010) propose that
‘might/can’ have the same lexical presupposition as ‘must’ (this also preserves the duality of ‘might/can’
and ‘must’). But, as Ippolito (2016) points out, ‘might/can’ simply do not seem to have an evidential
presupposition along these lines. This is clearest when we consider ‘might/can’ under attitude predicates.
On standard theories of presupposition projection, if ϕ presupposes ψ, then pS believes ϕq presupposes
pS believes ψq.26 If ‘might’, ‘can’, and ‘must’ carry an indirectness presupposition, then, it follows that
pS believes [must/might/can] ϕq will presuppose that S believes that the kernel for the ‘must/might/can’
does not directly settle [[ϕ]]. What is the kernel for an epistemic modal under ‘S believes’? A natural first
thought is that it is just S’s direct evidence. On this assumption − and given that presuppositions project
through negation and the antecedents of conditionals − the presuppositional approach thus predicts that
(22) and (23) presuppose that Susie believes her direct evidence does not directly settle whether John is
Pragmatic Strength. Pragmatic Strength is consistent with a weaker semantics, on which pMust ϕq entails ϕ but not knowledge of JϕK; or
with an account on which pMust ϕq dynamically entails ϕ, in the sense that accepting pMust ϕq commits you to accepting JϕK, but pMust
ϕq does not semantically entail ϕ (the account I will rely on in §4 has this structure). Lassiter (2016) and Giannakidou and Mari (2016)
argue that pMust ϕq does not entail knowledge of JϕK; these data, again, challenge (one part of) Semantic Strength, not Pragmatic Strength.
Lassiter for his part remains agnostic about whether pMust ϕq entails ϕ, but seems to accept Pragmatic Strength, insofar as he adopts a
lexical rather than pragmatic derivation of Indirectness.
22 And since endorsed and elaborated in Kratzer (2012a), Matthewson (2015), Lassiter (2016).
23 See Willett (1988b) on taxonomies of evidentials. Matthewson (2015) attempts to answer this challenge; see Sherman (2016) for a critical
response.
24 See Sherman (2016) again on this last point, as well as for an alternative approach to Indirectness which aims to avoid these problems. I
will not try to evaluate Sherman’s approach here, but I note that it insofar as it claims that intuitions about ‘must’ boil down to intuitions
about a question being left open, it may have something in common with my hypothesis that these intuitions boil down to intuitions about
redundancy.
25 von Fintel and Gillies (2010) and Lassiter (2016) also discuss alternate implementations, but the differences won’t matter for our purposes.
26 See Karttunen (1974), Heim (1992) and following. This prediction is uncontroversial, though some have argued that pS believes ϕq also
presupposes ψ when ϕ does.
13
the murderer.
(22)
Susie doesn’t believe that John might be the murderer.
(23)
If Susie thinks John might be the murderer, we should question him.
But, as Ippolito observes, this is wrong. For instance, we could felicitously follow (22) with ‘Indeed,
Susie thinks he’s not the murderer, because she knows that she saw someone else commit the murder’,
or (23) with ‘Indeed, for all we know, Susie knows that she saw him commit the murder and on that
basis thinks he is the murderer!’. Both these follow-ups, however, clearly conflict with the predicted
presupposition that Susie believes her evidence vis-à-vis the prejacent is indirect, and yet neither yields
infelicity. This suggests there is no such presupposition.
There is room for maneuver here by relaxing the assumption that the kernel for ‘must’ under ‘S
believes’ is supplied by S’s direct evidence.27 But whose direct evidence should it be instead? It is not
clear there are any good options. Suppose Susie has the Truman Show delusion: she thinks that everyone
on earth is watching her every action, and can see everything she sees. Moreover, let’s say that she thinks
this is common knowledge among everyone on earth. We can still assert (22), followed by ‘Indeed,
Susie thinks he’s not the murderer, because she knows she saw someone else commit the murder’. But
if Susie saw someone else commit the murder, then she thinks everyone also saw someone else commit
the murder (and that this is common knowledge among them) − and thus there is no actual person
whose kernel she believes not to directly settle whether John committed the murder. So what supplies
the kernel for ‘must’? It can’t be the direct evidence of any single actual person, or the distributed or
common evidence of any group of actual people, since Susie believes all those kernels directly settle the
prejacent. The possibilities seem limited.28
Both parts of the second strategy, then − reducing Support to Indirectness and then giving an independent explanation of Indirectness − face substantial open challenges. By contrast, I will argue now
that the third strategy − deriving Indirectness from Support, and then giving an independent explanation
of the latter − provides a satisfying route to solve Karttunen’s Problem.
3.2
From Support to Indirectness
Although Indirectness and Support don’t seem to be closely connected, I will argue that there is a pragmatic derivation from the latter to the former that goes by way of judgments about redundancy. In the
rest of this section I will explain that pragmatic derivation, before turning, in the next section, to an
explanation of Support.
In brief, the derivation is as follows. Pragmatic Strength says that an assertion of pMust ϕq is a bid
to update the common ground with JϕK; Support says that it is a proposal to do so on the basis of an
argument Γ. General principles forbidding redundant assertions entail that JϕK should not follow from
27 Following
broadly the strategy in von Fintel and Gillies (2011). Thanks to Kai von Fintel for discussion.
different approach is to treat these as cases of ‘local accommodation’ (Heim, 1983). But local accommodation is typically used to explain
away counterexamples to otherwise robust patterns; whereas I do not think there is any initial tendency to take away from either (22) or (23)
that Susie believes the salient evidence vis-à-vis the prejacent is indirect.
28 A
14
Γ in a way that is mutually recognized to be obvious. Finally, speakers are generally obligated to give
their best argument for JϕK if they’re giving an argument for JϕK at all. It follows that, in order for an
assertion of pMust ϕq to be felicitous, JϕK should not follow in a mutually obvious way from the best
argument a speaker of pMust ϕq can have for JϕK.
The first step in our derivation is the assumption, briefly defended in §3.1 above, that an assertion
of pMust ϕq is pragmatically strong in the sense that it is just as strong as an assertion of ϕ. Following
Stalnaker (1978), I assume that an assertion of ϕ is a proposal to update the common ground with JϕK.
Thus Pragmatic Strength says that an assertion of pMust ϕq is, inter alia, a proposal to update the
common ground with [[ϕ]].
The second step is to note that in general, when a speaker gives an argument in support of JϕK with
the intention of getting her interlocutors to accept JϕK on the basis of that argument, the argument in
favor of JϕK must be non-redundant in some sense. Compare the two variants in each of (24) and (25):
(24)
a. I put Patch in her box this morning, and no one has let her out. So she’s in her box.
b. ??I see Patch in her box. So she’s in her box.
(25)
a.
Clinton has amassed a majority of pledged delegates and superdelegates. So a woman will
clinch the Democratic nomination!
b. ??Clinton will clinch the Democratic nomination. So a woman will clinch the Democratic
nomination!
(24-b) strikes me as objectionable; there is something pedantic or redundant about it. Likewise for (25-b).
By contrast, (24-a) and (25-a) are fine. The difference seems to be that in (24-a) and (25-a), there is
enough epistemic space left between the argument in the first sentence and its conclusion in the second that its conclusion is not felt to be redundant. This intuition can be regimented as a norm against
redundant assertions, along the following lines:
Non-Redundancy: A proposal to update the common ground with JϕK on the basis of an
argument Γ is infelicitous if JϕK follows from Γ in a way that is mutually recognized to be
obvious.
In the next section I will further explore the justification for Non-Redundancy, show that it follows from
close variants on widely accepted norms, and say more about what it amounts to for JϕK to follow from Γ
in a way mutually recognized to be obvious. Note for the present, however, that Non-Redundancy nicely
captures the contrast between (24-a) and (24-b). The first is acceptable, since having put Patch in her box
in the morning, together with no one else having let her out, does not, in an intuitive sense, obviously
entail that Patch is in the box. The second is not, since it does follow in a mutually obvious way from
seeing Patch in her box that she is in her box.29
29 Note
that Non-Redundancy does not forbid post hoc support for an assertion with a redundant argument; it is perfectly fine to justify oneself,
if challenged, with ‘Because I saw it’. What Non-Redundancy forbids is making an initial bid to update the common ground with something
on the basis of an argument from which it follows in a mutually obvious way.
15
The last step in our derivation says that a speaker must give the best argument for JϕK that she has,
if she’s giving an argument for JϕK at all. To see the plausibility of this constraint, consider (26):
(26)
[John was at the Red Sox game and knows on this basis who won. He also read about the game
in the Boston Globe.]
a.
b.
[Max:] Who won the game?
[John:] ?? The Red Sox, according to the Globe.
If John intends (26-b) to answer Max’s question, then there is something strange about it; we expect
John to give his strongest evidence for the claim that the Red Sox won. In general, speakers are required
to share the best piece of evidence they have for a claim, if they are sharing evidence at all. This follows
naturally from a broadly Gricean vantage point on conversational dynamics. In (26-b), John is violating
Grice’s Maxim of Quantity by failing to ‘make his contribution as informative as is required (for the
current purposes of the exchange)’ (Grice, 1989). More precisely, the general lesson of cases like this is
a corollary of the Maxim of Quantity which I call Strongest Evidence:30
Strongest Evidence: When a speaker aims to update the common ground with JϕK on the
basis of an argument Γ, she is obligated to do so by providing the strongest argument − the
best piece of evidence − which she has for that claim.
We can now put these pieces together to derive Indirectness from Support. Support says that an
assertion of pMust ϕq is felicitous only if there is a shared argument for JϕK. Pragmatic Strength says
that an assertion of pMust ϕq is a proposal to update the common ground with JϕK. I will make the
plausible further assumption that an assertion of pMust ϕq is thus a proposal to update the common
ground with JϕK on the basis of a shared argument for JϕK (this is an assumption that will fall out of the
derivation of Support below). According to Non-Redundancy, JϕK must not follow from that argument
in a mutually obvious way. According to Strongest Evidence, that argument must constitute the best
evidence the speaker has for JϕK. It follows that in order for a speaker to be able to felicitously assert
pMust ϕq, JϕK cannot follow in a mutually obvious way from the speaker’s best piece of evidence
for JϕK. In other words, the speaker’s best evidence for JϕK must be indirect, in whatever sense of
indirectness is relevant to evaluating whether an argument is felt to be redundant.
Put differently: if a speaker has direct evidence (in the sense relevant to judgments about redundancy)
for JϕK, then, if she were to assert pMust ϕq, then due to Strongest Evidence and Support, she would
have to give that evidence as an argument on the basis of which she is proposing her interlocutors accept
JϕK; but then she will be bound to violate Non-Redundancy. So if she has direct evidence for JϕK, she
cannot assert pMust ϕq.
In sum: asserting pMust ϕq, the speaker has to ensure there is a shared argument which represents
her best evidence for JϕK, and yet is not so strong that it makes the ‘must’-claim sound redundant. Thus
30 See
Faller (2012) for more careful discussion of how this kind of reasoning would go. To spell out Strongest Evidence in more detail, we
need to be able to access a scale of evidential strength, according to which, say, direct perceptual evidence counts as stronger than any kind
of testimonial evidence − on this point see also Faller (2001). A norm like Strongest Evidence is related to but orthogonal to knowledge
norms on assertion of the kind discussed e.g. in Williamson (2000).
16
JϕK can’t follow in a mutually obvious way from her best evidence for JϕK. No parallel constraint follows
for non-modal claims − since Support requires only that ‘must’-claims be supported by an argument −
and thus Support, plus Pragmatic Strength, Non-Redundancy, and Strongest Evidence, entail a form of
Indirectness.
3.3
Non-Redundancy
This completes my derivation of Indirectness from Support. In the next section I will turn to the question
of how to predict Support. Before doing so, however, I will say more about the conversational architecture that underlies Non-Redundancy (in this subsection), as well as the predictions made by the present
derivation of Indirectness (in the next subsection).
Much more could be said about each of the assumptions I made in the last section, but NonRedundancy is the one most in need of further explanation here, because a central upshot of my account
is that the signal of indirectness associated with ‘must’ reduces to judgments about redundancy (rather
than to any of the categories of evidence standardly encoded by evidential marking). Saying more about
redundancy will help us evaluate this claim.
Non-Redundancy, again, runs as follows:
Non-Redundancy: A proposal to update the common ground with JϕK on the basis of an
argument Γ is infelicitous if JϕK follows from Γ in a way mutually recognized to be obvious.
Non-Redundancy follows from two premises. First, proposing to update the common ground with JϕK
on the basis of an argument Γ just is a proposal to update the common ground with Γ and then with JϕK.
Second, one should not propose a series of updates if, should each they all be accepted, the final update
will be judged to be redundant.
To spell out this simple idea, we need to investigate more general principles forbidding redundant
assertions. The commonly accepted principle forbidding redundant assertions is typically formulated
along the following lines:
Common Ground Entailment: Don’t assert ϕ if JϕK is entailed by the common ground.31
This norm is very natural if we think of conversations as cooperative enterprises whose goal is information transfer. Given our limited cognitive and temporal resources, we should not say what is already
common ground. There are, to be sure, felicitous cases in which speakers violate Common Ground Entailment, like cases of small-talk, as Stalnaker (1974) discusses. But, as Stalnaker argues, those seem not
to be counterexamples to Common Ground Entailment, but rather to be cases where Common Ground
Entailment can be violated − or where the common ground is being treated as smaller than it actually is
− for the sake of achieving a social goal.
Common Ground Entailment, however, is not quite right, even when we set aside small-talk cases. It
overgenerates in cases like (27):
31 See
Stalnaker (1974, 1978, 1973). Some version of this norm applies at the subsentential level as well, which plays an important role for
embeddings of ‘must’; see e.g. Schlenker (2008, 2009), Mayr and Romoli (2016).
17
(27)
a.
b.
c.
If it’s raining out, then Bob brought his umbrella. If Bob brought his umbrella, then he
won’t have noticed that we had the roof redone. And, it was raining out.
Ok.
So Bob won’t have noticed that we had the roof redone.
Common Ground Entailment wrongly predicts that (27-c) will be infelicitous.32 (27-c) is acceptable,
however: it simply makes explicit the conclusion of a somewhat complex chain of reasoning, one which
the speaker would not necessarily have expected her interlocutor to draw on her own.
The problem, intuitively, is that entailment is too strong of a notion to play the role it does in Common
Ground Entailment. Instead, I propose to treat an assertion as redundant if it is entailed by the common
ground in a way mutually recognized to be obvious − obvious in a derogatory sense: a sense which
makes an assertion superfluous (there is a sense of ‘obvious’ in which it is acceptable to assert obvious
things; that is not the sense in question here). In the present case, we say that (27-c) is not a mutually
obvious entailment of (27-a), and thus we rightly predict its felicity. What counts as a mutually obvious
entailment will depend on the context, and will be a matter of ongoing negotiation; I discuss this point
further below (see example (30)).33
Merely changing ‘entailed’ to ‘entailed in a mutually obvious way’ in Common Ground Entailment
does not yet, however, yield a satisfying principle. The modified principle will undergenerate in important cases. Consider:
(28)
a. What time is the movie?
b. The cinema website says that it’s at 7:30.
c. Ok.
d. ??So the movie’s at 7:30.
(28-d) is, in most contexts, unacceptable − intuitively, because it is felt to be redundant. But it is not
as though (28-b) entails (28-d). To capture data like these, we need to use a more permissive relation
than entailment: a relation which can capture the sense in which ‘the movie is at 7:30’ follows from ‘the
cinema website says that the movie is at 7:30’.34 Here I will simply use ‘follows from’ to capture this
notion, with the understanding that we will get a feel for the notion from examples like (28). If Γ entails
JϕK, then JϕK follows from Γ; but JϕK also might follow from Γ if JϕK is a reasonable default inference,
or is highly probabilified by Γ. Much more needs to be said here, but this suffices for our purpose.
The most natural way to capture these two revisions to Common Ground Entailment is to model the
common ground as a set of propositions which is closed under mutually obvious inference, but which is
not closed under logical entailment in general. Then we can state our redundancy principle as follows:
32 It
is possible to respond that (27-c) is an acceptable violation of the norm, rather than that it shows the norm to be too strong. This is hard
to adjudicate, but it doesn’t matter for our purposes: we could go this way and adopt a version of Non-Redundancy which, likewise, is put in
terms of entailment, with the caveat that the principle can be violated in many cases.
33 See Barker (2009) and Kripke (2009) for related discussion, as well as literature on the normativity of logic (e.g. MacFarlane (2004)). As
Justin Khoo has pointed out to me, this will help explain why it can be rude to spell things out in detail: this shows that you are not treating
something as mutually obvious which your interlocutor assumed was.
34 A different response to data like this would be to hold that in cases like this, we accommodate an enthymematic premise, so that this really
is an entailment. I do not see how to choose between this response and the present one, and thus would accept either formulation.
18
Common Ground Settlement: Don’t assert ϕ if JϕK is in the common ground.
From a formal point of view, there is no difficulty in treating the common ground as not closed under
logical entailment, but closed under mutually obvious inference. The common ground is simply a set of
propositions. More needs to be said, of course, about what distinguishes a proposition which is common
ground from one which is merely entailed by the common ground. I take it that this distinction will track
something about behavioral dispositions.
Note an important upshot of the present considerations about redundancy. The common ground is
standardly thought to be entirely derivative from individual mental states.35 If the present considerations
provide a convincing argument in favor of modeling the common ground as a set of propositions closed
under obvious inference but not under logical entailment, then these considerations can be extended
to argue in favor of likewise modeling individual attitudes as sets of propositions with these closure
properties.36 I will leave exploration of this point for future work; here I will emphasize a broader
methodological note illustrated by this discussion: intuitions about assertability, and in particular about
redundancy, may give us an important source of new data about the structure of mental states.
There is, again, much more to be said about Common Ground Settlement − about when something
follows from a set of premises, when it does so in a way mutually recognized to be obvious, and what
the underlying theory of mind should be to account for this. I leave further exploration of these topics
for future work.37 What is important for present purposes is that Non-Redundancy is just a corollary of
Common Ground Settlement. If a speaker proposes to update the common ground with JϕK on the basis
of an argument Γ, then she intends her interlocutors to first accept the elements of Γ and then − on
that basis − accept JϕK. In other words, she intends the elements of Γ to become common ground, and
then for JϕK to become common ground. In order for an action of this kind to be acceptable, according
to Common Ground Settlement, it must be the case that JϕK does not follow from Γ in a way mutually
recognized to be obvious.
3.4
Predictions
The present proposal derives Indirectness from general principles about redundant assertions, and thus
makes a striking empirical prediction: namely, that S’s evidence Γ for JϕK counts as indirect in the sense
relevant to Indirectness just in case an assertion of ϕ following sequential assertions of the elements
of Γ does not strike us as redundant. This provides a new answer to the question of how to spell out
35 See
e.g. Stalnaker (2002).
appropriate semantics for attitude reports would thus be a neighborhood semantics, rather than a standard modal semantics (Montague
(1970), Scott (1970)). See Kratzer (2012b, 19-20) for a general defense of representing attitudes with sets of propositions, rather than simply
sets of possible worlds: as Kratzer puts it, ‘Representing the content of recommendations, claims, beliefs, orders, wishes, etc. as premise
sets. . .offers us the priceless opportunity to represent connections between propositions in a given premise set. The content of such speech
acts and attitudes can now be seen to have an inherent structure that encodes which propositions stand and fall together under challenge.
This structure is lost if information contents are directly represented as sets of possible worlds.’ Note that treating the common ground as not
logically closed by no means commits us to linking the propositions in the common ground with particular linguistic acts: while linguistic
acts are one way in which propositions can enter the common ground, there are lots of other ways they can do so as well. An alternate
approach would be to model the common ground, and thus mental states in general, as overlaid with a partition (see Yalcin (2011, 2016a),
Bledin and Rawlins (2016)). This approach accomplishes much the same goal as the present one but is more constrained than the present
approach. I will not try to decide between them here.
37 One possibility is to adapt one of the proposals for formalizing Indirectness given in von Fintel and Gillies (2010).
36 The
19
the notion of indirectness involved in Indirectness: an answer which I will now argue provides a better
characterization of the data than the alternative view, according to which ‘must’ lexically encodes a
requirement that the speaker’s evidence be indirect in a sense that lines up fairly closely with intuitions
about whether evidence is direct or indirect, and with categories which are encoded by grammatical
evidentials (I’ll call this ‘an evidential approach’; von Fintel and Gillies (2010) is the clearest example
of this kind of approach).
I will highlight a few points. First, the present approach predicts that ‘must’-claims based on reliable
testimony like (29) will not be acceptable:
(29) ??The website says the movie is at 7:30. So the movie must at 7:30.
Our approach predicts this, since reliable testimony for [[ϕ]] is typically felt to be a redundant argument
for [[ϕ]], as shown by examples like (28) (a non-modal variant of (29)).38 By contrast, this is surprising
on an evidential approach, since testimony is, intuitively, indirect evidence (it is natural to say that you
know that the movie is at 7:30, but that you know indirectly, via the website). An evidential approach
thus must simply stipulate that testimony ‘counts as direct’ for the purposes of evaluating ‘must’.
The second prediction of our approach worth highlighting is that what counts as redundant in a given
context − and thus judgments about the felicity of ‘must’ − depends on what counts as mutually obvious
in that context. Thus, e.g., while (29) is infelicitous out of the blue, it may be felicitous in a context in
which the inference from website listings to fact is not generally accepted, as in (30):
(30)
Google says that the movie is at 7:30. Websites listing movie times are generally extremely
unreliable. Google is extremely reliable, though, so the movie is indeed at 7:30.
Given the felicity of (30), we predict that a ‘must’-claim will be felicitous here as well; and indeed, (31)
is felicitous:
(31)
Google says that the movie is at 7:30. Websites listing movie times are generally extremely
unreliable. Google is extremely reliable, though, so the movie must indeed be at 7:30.
More generally, we rightly predict that judgments about the felicity of ‘must’-claims depend on what
counts as mutually obvious in context. It is not as clear how an evidential approach would predict this,
since it does not seem like what counts intuitively as direct versus indirect evidence varies from context
to context: our evidence for the time of the movie is equally indirect, in an intuitive sense, in (31) and in
(28).
Third, we can explain why ‘must’-claims that conclude a complicated argument are generally acceptable, even if the premises of the argument entail its conclusion. Examples of this kind, in particular those
involving mathematical or logical claims, are the most puzzling examples for an evidential approach to
‘must’. ‘Must’ is often warranted in mathematical or logical contexts, like (32).39
38 Why
sequences like this are treated as redundant is, of course, an important question for theories of redundancy to address, but one I will not
answer here.
39 Is the ‘must’ here epistemic? Some have argued that this ‘logical’ ‘must’ is not genuinely epistemic (e.g. Giannakidou and Mari (2016),
20
(32)
If the set of validities were decidable, then the halting problem would be decidable. The halting
problem is not decidable. So the set of validities must be undecidable.
It is not clear what the evidential approach would predict about (32). It is not clear that there is an
intuitive sense on which our evidence that the set of validities is undecidable is indirect. Perhaps the
evidential approach would claim that evidence for pure mathematical claims is always indirect in the
relevant sense. But in addition to being somewhat stipulative, this response runs into trouble when it
comes to examples like (33):
(33) ??24 plus 24 must be 48.
If all evidence from pure mathematics is indirect, then the evidential approach will wrongly predict that
‘must’ is warranted in (33) and sentences like it. It is, again, not clear how the evidential approach can
distinguish between cases like (32) versus (33). By contrast, our approach can. The conclusion of (32)
does not follow in a way that is mutually obvious from the premises, whereas the conclusion of (33)
does.
These points confirm the key claim of the present approach to Indirectness: what matters for determining whether a ‘must’ is warranted is not whether the speaker’s evidence for it is indirect in any
intuitive sense, but rather whether the speaker’s evidence for it makes the prejacent mutually obvious.40
I will conclude this section by discussing a different prediction of the present account. I have proposed that Indirectness arises due to conversational norms. It is a hallmark of pragmatic phenomena like
this that they can be cancelled, since the underlying conversational norms are generally defeasible. We
thus predict that Indirectness will be cancelled when one of the underlying norms is not in play. This
prediction, again, is borne out, in particular in contexts in which Strongest Evidence is not in play because it is overridden by considerations which prevent the speaker from sharing her strongest evidence
for [[ϕ]]. For instance, suppose that Mary is at Tom’s party. She goes out to the street to smoke, where she
runs into Ben. She knows Ben wasn’t invited to the party, and doesn’t want him to know that she was
invited. Ben can hear music coming from Tom’s place, and asks Mary what’s going on at Tom’s. Mary
wants to communicate that he’s having a party, but she doesn’t want to share her strongest evidence for
this − and doesn’t seem to be under any obligation to do so, since she is trying not to hurt Ben’s feelings.
In this context, she can felicitously assert (34):
Goodhue (2016)). While I cannot say anything to decisively rule out that possibility, I note two things that militate against this option
and in favor of a unified approach: first, it is inelegant to multiply modal flavors further than we need to. Second, even if we say that the
logical ‘must’ is not epistemic, we still need a theory of its distribution, since it is not always warranted, even when its complement is a
logical consequence of the common ground (as examples like (33) show); simply saying that this ‘must’ is logical, not epistemic, thus does
not yet explain its behavior. My account avoids multiplying categories of modality, and explains the distribution of ‘must’ in this kind of
environment.
40 That said, the present approach does not immediately explain all facts about the distribution of ‘must’ in mathematical environments. For
instance, it seems to me that ‘must’ is never warranted in environments just involving simple applications of arithmetic operations, even when
the conclusion of the argument in question is not at all obvious, as when it is the conclusion of a difficult sum. One possible, if somewhat
stipulative, explanation for this is by way of the discussion of weakness in §6: arithmetic is the kind of thing that we expect speakers to
put their full assertoric authority behind, rather than to pass off to the group − except perhaps in certain pedagogical environments. This
explanation may extend to another class of cases, brought to my attention by Agnes Callard, involving moral judgments. ‘Lee was late, and
therefore disrespectful’ is felicitous, but ‘Lee was late, and therefore must be disrespectful’ is very strange. The explanation, again, may be
that we expect speakers to put the full weight of their authority behind assertions like this.
21
(34)
Given the music, it must be some kind of party.
(34) may be misleading, but it strikes me as perfectly felicitous, despite the fact that Mary’s evidence
that it’s a party is direct. The prediction of our pragmatic account is thus borne out: Indirectness can be
violated when one of the underlying pragmatic norms can itself be appropriately ignored.41
4
Support
Support plus independently motivated pragmatic principles thus provide a theoretically and empirically
satisfying explanation of Indirectness. I turn now to the question of how to account for Support. I briefly
survey and criticize the few extant proposals before giving my own account.
4.1
Extant Proposals
Support says that ‘must’ requires that an argument for its prejacent be made salient. A natural first
thought about how to predict Support is to treat ‘must’ as containing something like an implicit indexical
for an argument: ‘must’ means roughly ‘it follows from this evidence that. . .’, where the implicit ‘this’,
just like an overt ‘this’, requires a salient referent. Stone (1994) suggests an account along just these
lines: on his approach, ‘must’ has a lexical argument place which must be saturated by an argument made
salient by context. In other words, ‘must’ is a two place operator, taking an argument and a proposition
p, which says that the argument provides ‘a decisive reason to adopt the belief that p’.
A solution along these lines, natural though it is, has two drawbacks: one minor, one which strikes
me as decisive. The minor drawback is that − like von Fintel and Gillies (2010)’s lexical stipulation of
Indirectness − this does not look like a very satisfying explanation of Support. Support is not a quirk
of English ‘must’, but rather a cross-linguistically robust feature of anything that has the meaning of a
strong epistemic necessity modal. This fact is not well explained if we predict it by lexical stipulation.
A much more substantial issue stems from the fact that no parallel to Support shows up for unembedded epistemic possibility modals. First, note that e.g. (35):
(35)
[Julie’s cat has been sneezing a lot lately. Ben asks her how the cat is doing. Julie says:]
a.
Not so great. I need to take him to the vet actually, he might have an upper respiratory
infection.
b. Not so great. I need to take him to the vet actually, he has an upper respiratory infection.
c. ??Not so great. I need to take him to the vet actually, he must have an upper respiratory
infection.
Suppose the conversation ends here. As Support predicts, (35-c) is infelicitous as it stands, without an
argument. If we adopted a corollary of Support for ‘might’ or ‘can’, then we would wrongly predict
41 It
is less clear to me whether similar cases can be constructed in which it is Non-Redundancy which is suspended, since Non-Redundancy
already has an element of context-sensitivity built in (since what counts as ‘mutually obvious’ − in the objectionable sense relevant here −
is itself context-dependent). But if we can find contexts in which it is suspended, then we predict that in those contexts as well, Indirectness
will be suspended.
22
(35-a) to be equally infelicitous. But (35-a) − like the non-modal variant in (35-b) − is perfectly fine
here. ‘Might’ thus does not seem to be subject to a Support-like constraint.
If we took Stone’s approach, however, then, provided we assume that ‘must’ and ‘might/can’ are
duals, we would predict that ‘might’ has an anaphoric requirement for an argument, just as ‘must’ does:
if ‘might’ means ‘not must not’, then the argument requirement of ‘must’ will project through negation,
and thus ‘might’ will require a salient argument, too. We could avoid this by giving up the assumption
that ‘must’ and ‘might/can’ are duals, and that ‘might/can’ does not have a lexical argument place for an
argument. But, crucially, going this way leads to a serious new puzzle. Assuming we treat ‘cannot’ as
‘not (can)’, then we will predict that unembedded ‘cannot’ does not have an anaphoric requirement for
an argument any more than unembedded ‘can’ does. But this is wrong: the same examples we used to
motivate Support for ‘must’ above can be used to motivate it for unembedded ‘cannot’ (modulo obvious
changes). Thus, for instance, consider (36):
(36)
Emma notices that her neighbor Phil hasn’t taken in his mail in some time, and concludes that
he is out of town. Another neighbor asks if Phil is around. Emma responds:
a.
b.
c.
d.
No, he can’t be.
No, he’s not.
No, he can’t be: no one has taken his mail in for a week.
No, he’s not: no one has taken his mail in for a week.
The exchange ends here.
(36-a) is marked as compared with the other variants in (36). As with the examples involving ‘must’,
then, ‘can’t’ seems to require that an argument for its prejacent be made salient.
Thus a lexical derivation of Support along the lines Stone suggests faces a dilemma: either treat
‘must’ and ‘might/can’ as duals, and wrongly predict that the latter have a Support-like requirement;
or do not treat them as duals, and wrongly predict that ‘cannot’ lacks a Support-like requirement. This
approach thus strikes me as a non-starter.
Similar criticisms extend to the account suggested in Swanson (2015), who builds on Kratzer (1981)
in adopting a premise semantics for epistemic modals, with the added requirement that those premises
be publicly available.42 Setting aside the question of where exactly this latter requirement comes from,
it looks like, again, this approach faces a dilemma. If we treat ‘might/can’ as the duals of ‘must’, then
this explanation will overgenerate: it will wrongly predict that they are likewise subject to Support, since
they will likewise require a set of premises to be made public. Alternately, we could abandon duality,
but then we cannot explain Support for ‘cannot’.43
42 Moss
(2015) gives a proposal which is relevantly similar to Swanson’s for our purposes.
alternative, suggested to me by Eric Swanson (p.c.), maintains duality, but treats the salience of the argument in question as part of the
asserted content of ‘must’: pMust ϕq means pThere is a salient argument Γ which entails [[ϕ]]q; and so pMight ϕq means pThere is no
salient argument Γ which entails [[¬ϕ]]q. While this does avoid the present objection, it seems to me to make the wrong prediction about the
meaning of ‘might’; it predicts that the absence of any salient arguments entails the truth of all ‘might’ claims. But this seems false; when I
start a new conversation, for instance, I cannot assert pMight ϕq for any ϕ just because there are no salient arguments in the context. Perhaps
this could be explained pragmatically, though, or by adopting a threshold credence requirement along the lines of Swanson (2015).
43 An
23
The present objection can also be extended to a treatment of Support as a presupposition.44 Presuppositions project through negation; thus a presuppositional approach would either treat ‘might/can’ as
duals of ‘must’, and thus wrongly predict that ‘might/can’ obey Support; or would abandon duality, and
once again fail to predict Support for ‘cannot’.
4.2 Support as a Manner Implicature
We can avoid these problems by deriving Support as a pragmatic implicature, along the following lines.
The derivation depends in part on adopting a semantics for ‘must’ defended in Stalnaker (2014) and
Mandelkern (2016a); I will not try to motivate the semantics here in general terms.45 According to that
semantics, ‘must’ is a universal quantifier over the set of worlds compatible with what is common ground
after the ‘must’-claim in question has been made and accepted or rejected − call this the ‘prospective
common ground’. pMust ϕq thus means, roughly, pWe will commonly believe JϕK after this claim is
made and assessedq; ‘might’ is treated as the dual of ‘must’, and thus pMight ϕq will mean pJϕK is
compatible with what we commonly believe, after this claim is made and assessedq.46 The basic idea is
the familiar one that ‘might’ and ‘must’ are used to coordinate on what structural properties the context
set has.47 I adopt the present semantics for ‘must’ partly for concreteness, partly because I think it is
plausible, and partly because it lends itself naturally to the present derivation. But as will become clear,
a similar derivation of Support may well be possible with a different underlying semantics for ‘must’:
the crucial features of the semantics for present purposes are, first, that it is pragmatically strong; and,
second, that it makes salient the question of the interlocutors’ collective doxastic relationship to its
prejacent. Any semantics for ‘must’ with these two features will suffice for present purposes.
On this approach, pMust ϕq thus has the basic update effect of adding JϕK to the common ground:
in other words, coming to accept pMust ϕq has the essential effect of coming to accept [[ϕ]]. pMust ϕq is
thus in competition with a different assertion which has the same essential update effect but which is
structurally simpler: namely, an assertion of ϕ alone.48 Because ϕ is structurally simpler,49 choosing
an assertion like pMust ϕq instead requires some explanation. Because the two options have the same
update effect, the interlocutors cannot reason that the speaker chose one of them because she knew
the other was false (as in scalar reasoning). There is, however, an important difference between pMust
44 A
suggestion due to Eric Swanson (p.c.).
approach and Stalnaker’s differ in important ways, but the differences do not matter for our purposes.
46 This gloss is only rough for two reasons. First, because we might go in for either a factive or a non-factive notion of common ground in
spelling out this semantics. I will prescind from deciding this question, which is orthogonal to my interests here. Second, I do not claim
that ‘might’ and ‘must’ work in exactly the same way as its gloss; there are important differences, in particular with regard to how these
constructions embed and how they interact with conversational norms. See Mandelkern (2016a) for further discussion. The basic idea for
embeddings is that the domain of quantification for embedded ‘must’ is determined by its local context. Thus pS believes must ϕq does not
mean that S believes the prospective common ground will entail ϕ, but rather that S believes her evidence entails ϕ.
47 This idea goes back to Stalnaker (1970) and has been taken up in much of the subsequent literature on epistemic modals.
48 See Degen et al. (2015) for a different manner-implicature based approach to explaining the behavior of ‘must’. Like the present approach,
that approach relies on the assumption that pMust ϕq is a costly alternative to ϕ whose use must be somehow explained. In contrast with the
present approach, that approach attempts to derive Indirectness directly, rather than by way of Support, and without adverting to the different
QUDs raised by modal vs. non-modal variants.
49 I will not spell out assumptions about how we calculate which alternatives are relevant. It seems fairly plausible that however we do so, ϕ
will count as a relevant alternative to pMust ϕq (and ¬ϕ as a relevant alternative to pCan’t ϕq); this follows e.g. on the account given in
Katzir (2007), according to which alternatives are calculated by the deletion or replacement of nodes at LF (modulo issues about the change
from a non-finite verb in ϕ when it is embedded under ‘must’ to a finite verb when it is not embedded).
45 My
24
ϕq versus ϕ alone: namely, that the former makes salient the question of the interlocutors’ collective
doxastic relation to JϕK, while the latter does not: it is only about JϕK. There are a variety of ways to
make this intuition precise. It doesn’t matter for our purposes how we choose between them.50 For our
purposes, we can just treat the question raised by an assertion of [[ϕ]] as the two cell partition {ϕ, ϕ̄}.
Then the question made salient by an assertion of JϕK is just the question whether JϕK is true; whereas
the question made salient by an assertion of pMust ϕq is the question whether we will come to commonly
accept JϕK. Given that pMust ϕq has a simpler alternative with the same basic update effect (namely ϕ),
if a speaker chooses to use this more complex expressions, then we will seek an explanation of this fact.
Given that the chief difference between pMust ϕq and ϕ is in the question made salient by each, we will
therefore reason that, in choosing the more complex option, she wishes to raise to salience the question
of the group’s collective doxastic relation to JϕK. Given that she is, moreover, proposing that the group
come to accept JϕK, we can reason that she is proposing that the group accept [[ϕ]] on the basis of a
reason to believe [[ϕ]] which she is highlighting.
What kind of reason for accepting JϕK would be worth highlighting in this way? Whenever a speaker
proposes to update the common ground with JϕK, this provides ceteris paribus reason for her interlocutors to accept JϕK: namely, the speaker’s authority. A reason of this kind, therefore, is totally humdrum,
and thus not worth highlighting. The use of an assertion like pMust ϕq over an assertion of ϕ alone thus
can be justified only if the speaker wishes to highlight a substantial argument for JϕK − i.e., an argument over and above the speaker’s authority − that the speakers have in common, and on the basis of
which the speaker wishes her interlocutors to accept JϕK. In other words (the interlocutors will reason),
the speaker wishes them to accept JϕK on the basis of an argument that is commonly available to them.
She must, therefore, ensure that such an argument is salient − either by providing it, or being assured
that her interlocutors can recover it from the common ground, possibly by accommodation. In short,
then, Support arises as a manner implicature in the process of trying to understand why a speaker chose
a structurally more complex option, when a simpler alternative would have had the same basic update
effect. This provides an explanation of how Support arises for pMust ϕq.
This approach avoids the problems discussed for the alternate approaches to Support considered
above. First, this approach derives Support from a simple, independently motivated modal semantics,
rather than lexical stipulation, and thus explains why Support arises for anything that has the meaning
of ‘must’. Second, it allows us to treat ‘might/can’ as duals of ‘must’, since the derivation of a Support
constraint for ‘might/can’ will be blocked by the fact that assertions of pMight ϕq and ϕ do not have the
same basic update effect. A crucial step in the derivation, again, was that pMust ϕq has a structurally
simpler alternative with the same basic update effect. This is not true of pMight ϕq. Assuming that ‘can’
and ‘might’ mean the same thing, and that ‘not’ scopes over ‘can’ in ‘cannot’, the derivation given above
will extend immediately to unembedded ‘cannot’, and thus will explain why Support and Indirectness
arise for ‘cannot’ just as for ‘must’.51
50 See Lewis (1988), van Kuppevelt (1995), Ginzburg (1995a,b), Roberts (2012), Yablo (2014), Bledin and Rawlins (2016) and citations therein
for recent discussion of related topics.
Fintel and Gillies (2010) cast doubt on a pragmatic derivation of Indirectness which bears some similarity to the present approach: it
treats pMust ϕq and ϕ as competitors, with the latter stronger than the former. The problem, of course, as we have seen, is that the latter
51 von
25
Note that for the same reason that this derivation of Support is blocked for ‘might’, the corresponding
derivation of Support will likewise be blocked for weak epistemic necessity modals like ‘ought’ and
‘should’, as well as probability modals, since assertions of pought/should/probably ϕq and ϕ do not have
the same basic update effect (the former are uncontroversially weaker than the latter).52 The prediction
that ‘might’ does not carry a Support constraint seems correct, and is confirmed by examples like (35).
The corresponding prediction for ‘ought/should/probably’ likewise seems correct:
(37)
a.
b.
When do you want to meet?
Let’s say Thursday;
(i) I should be free then.
(ii) I ought to be free then.
(iii) I’ll probably be free then.
(iv) I must be free then.
If the conversation ends here, responses (37-b-i), (37-b-ii), and (37-b-iii) are all acceptable; by contrast,
(37-b-iv) is a strange way to end the conversation, and seems to require that some argument be given
(‘. . .my secretary always leaves my Thursdays open’). The prediction of the present account, then, that
modals which are not Pragmatically Strong, like ‘might/ought/should/probably’, do not carry a Support
constraint, thus seems correct.53
The derivation of Support is blocked for other modals. Is it ever cancelled in the way that, as we saw
above, Indirectness can be? Even though I have proposed that Support is a conversational implicature,
it is not clear to me that we thus predict that it can be cancelled. The pragmatic principle underlying
the derivation is something like pIf ϕ and ψ are alternatives with the same basic update effect, use the
structurally simpler one unless you have a good reason not toq. It is not clear that this principle is ever
suspended, and thus not clear to me that we predict that Support is ever cancelled. And indeed, this
seems to match the empirical picture: it is not obvious that Support ever is in fact cancelled.54
Finally, the present proposal makes an empirical prediction which is worth highlighting, since it
offers the best means for testing its plausibility: any construction which has the features which played
an essential role in the derivation is predicted to give rise to a Support-like constraint. That is, any
construction which has ϕ as a structurally simpler relevant alternative; which has the same basic update
effect as an assertion of ϕ; and which highlights the speakers’ collective doxastic relation to JϕK, is,
ceteris paribus, predicted to give rise to Support (and thus also Indirectness). Further research should
examine other expressions that share these three features to see if this prediction is borne out.55
is not stronger than the former. Like von Fintel and Gillies, I am giving a pragmatic derivation of Support which treats pMust ϕq and ϕ
as competitors. Crucially, though, we avoid the analogue to von Fintel and Gillies’ objection, because this derivation does not rely on the
assumption that one of these is stronger than the other: instead, it goes by way of considerations of complexity and what could merit choosing
the more complex expression.
52 Beyond this assumption, I will not take any further position on the semantics of ‘ought/should/probably’.
53 And again, the Indirectness inference for ‘might/ought/should/probably’ is easy to explain pragmatically without a detour through Support,
through the strategy discussed in §3.1, since these modals are pragmatically weak.
54 Modulo all the caveats above about reasons that we might be willing to ignore genuine violations of Support.
55 pIt’s agreed that ϕq, pLet’s agree that ϕq, pWe should believe ϕq, and pIt is clear that ϕq all seem to have these properties, do indeed appear
to be governed by corollaries of Support and Indirectness. This suggests that, even if the pragmatic derivation I have given here is mistaken
in some of its details, the explanation for both Indirectness and Support is pragmatic, and stems from the properties I have pointed to. See
26
5
Embeddings
This concludes my main line of response to Karttunen’s Problem. Before concluding, I take up two
residual questions. The first regards how to extend the present account to embeddings of ‘must’. First,
note that Indirectness persists when ‘must’-claims are embedded under connectives as in (38-b) and
(38-c), which communicate that the speaker’s evidence that Mark came to the party is suitably indirect:
(38)
a.
Who came to the party?
(i) John did, and Mark must have; they’re inseparable.
(ii) John did, and Mark must have.
(iii) John did and Mark did.
Second, Support also seems to be in play in cases like this. It is satisfied in (38-b), which is felicitous
as it stands. (38-c) is felicitous only if we can accommodate an argument that Mark came (for instance,
that John and Mark always go out together). By contrast, (38-d) requires no such inference: there need
be no salient argument that Mark came to the party.
A natural first response to cases like this is that, since these signals persist through embeddings, they
must be semantically encoded. But this would be too quick. On the one hand, it is not at all clear how
to capture these data semantically. As I discussed in §3.1 and §4.1, respectively, encoding Indirectness
or Support as a presupposition does not seem plausible, given that neither one projects in the way presuppositions are known to. Encoding them as conventional implicatures would not do any better, since
it would predict unconditional projection, and thus face the same counterexamples that make trouble
for the presuppositional approach.56 And it is clear that they are not part of the main content of ‘must’,
since e.g. pNot must ϕq clearly does not mean the same thing as pMy evidence for ϕ is direct, or fails
to entail ϕq (one cannot assert ‘It’s not the case that it must be raining out’ because one sees that it’s
raining out, except as a kind of meta-linguistic negation used in response to a ‘must’-claim which has
just been made). These considerations suggest that it is not at all obvious how a semantic approach to
these signals would work.
On the other hand, we can explain these cases pragmatically by taking essentially the same strategy
as in the unembedded cases. The details of this will depend on an account of how ‘must’ embeds; I cannot
get into the details of this here, but I will sketch the extension. The basic idea is that ‘must’ embedded
under connectives quantifies over the worlds compatible with the prospective common ground and the
local context of ‘must’.57 Thus, for instance, the ‘must’ in (38-a-i) quantifies over the worlds in the
prospective common ground where John came to the party. Then, note that (38-a-i) is in competition
with a structurally simpler non-modal variant, namely (38-a-iii), which omits the ‘must’ but which has
the same basic update effect. The difference in meaning between (38-a-i) and (38-a-iii) is that, while both
again Barker (2009); thanks to Kai von Fintel for discussion.
is so at least on the approach due to Potts (2005); others have however since resisted the claim that conventional implicatures project
unconditionally (e.g. Martin (2016)), which may create room for maneuver here.
57 On local contexts, see e.g. Stalnaker (1970), Karttunen (1974), Heim (1982). An interesting question which I do not have space to pursue
here is the difference between a conditional with the form pIf ϕ then ψq versus pIf ϕ then must ψq. The answer to this question will depend,
of course, on the semantics for the conditional without ‘must’.
56 This
27
communicate that Mark came to the party, the former, but not the latter, also highlights our collective
doxastic relation to the prejacent, in roughly the same way that unembedded ‘must’ does. Thus − because
the speaker chose the more complex variant − she must think that our collective doxastic relation to the
prejacent is worth highlighting, and thus that there is an argument commonly available for the prejacent
(at least once we take into account the information in its local context − in this case, that John came).
It is widely thought that Non-Redundancy holds not only at the level of sentences, but also at a local,
sub-sentential level (such a theory is required to explain the infelicity of a sentence like ‘If John is in
Paris, then John is in Paris and he’s having fun’).58 We can then reason as above to show that the relevant
argument must be indirect: the argument must be shared (by the time we reach the local context), and
thus the prejacent should not follow from it in an overly direct way, at risk of being locally redundant.
This approach extends straightforwardly to other embeddings of epistemic modals under connectives −
provided there exists a non-modal variant which communicates the same basic information.59
The indirectness of embeddings of ‘must’ under attitudes, like (39), can be explained in a simple but
slightly different way.
(39)
James thinks Bob must be in his office.
Here, again, the ‘must’ seems to suggest that James’ evidence is indirect. How to predict this will depend
on the interaction between attitude verbs and ‘must’. I take up this topic in detail in Mandelkern (2016b),
where, following accounts like those in e.g. Stephenson (2007a,b), I propose that attitude predicates like
‘thinks’ shift the modal base for ‘must’ to the set of worlds compatible with the evidence of the agent
of the attitude ascription: in other words, attitude verbs ‘anchor’ the modal base to the agent in question.
Thus an attitude ascription like (39) says that James believes his evidence entails that Bob is in his
office. A comment on what kind of evidence an agent thinks she has will, in turn, only generally be
worth making if that evidence is somehow worthy of note. As we saw above, evidence is generally only
felt to be worth drawing attention to in natural language if there is some kind of epistemic space left
between that evidence and its conclusion. Thus we can conclude from attitude ascriptions like (39) that
there is some interesting space between James’ evidence and the conclusion that Bob is in his office; in
other words, that James’ evidence is indirect in roughly the sense cashed out above.
More needs to be said to spell out these accounts in detail − in particular, a thorough account of
embedded epistemic modals must be spelled out. But this discussion should suffice to give a sense of
how the present account makes sense of the persistence of indirectness signals for embedded ‘must’.
6
Weakness
In this final section, I address how my account explains the felt weakness of some ‘must’-claims, and the
intuition that in some cases ‘must’ is obligatory.
I have assumed that ‘must’ is pragmatically strong in the sense that an assertion of pMust ϕq is a
proposal to update the common ground with JϕK. But our account nonetheless provides a distinctive
58 See
e.g. Schlenker (2009), Katzir and Singh (2013), Mayr and Romoli (2016) for discussion of theories of redundancy at the local level.
again, this approach will correctly not extend to ‘must’ under negation.
59 Thus,
28
and precise way of cashing out the intuition that ‘must’ is, in many cases, felt to weaken a claim.60 In
uttering pMust ϕq, on my account, a speaker proposes that her interlocutors accept JϕK on the basis of an
argument that is publicly available, rather than on her assertoric authority. The conversants’ acceptance
of JϕK thus becomes the responsibility of the whole group of speakers, rather than of the speaker alone.
pMust ϕq is thus weaker than ϕ in the way in which pWe should believe ϕ, on the basis of such and
such groundsq is: not because it results in a weaker update, but because it makes the responsibility for
updating with JϕK the responsibility of the whole group. By passing off the responsibility for the update
to the group, the speaker places the responsibility for accepting and evaluating JϕK wholly on the group
of conversants. She may thus be felt to be pointing out that her reasons for believing JϕK merit the
scrutiny of the group, and thus that they are not as strong as would be required for her to stake her own
authority on her assertion.
This perspective on the weakness of ‘must’ may help shed some light on cases of the kind discussed
in Ninan (2014), in which an assertion with ‘must’ is acceptable, but an assertion without ‘must’ is not.
Consider (40):
(40)
[Suppose A and B are friends with a couple, Carl and Diane, who have been dating for a long
time and are likely to get married at some point in the future. Suppose that, prior to the following dialogue, B has not heard any recent news concerning Carl and Diane’s relationship. Now
consider:]
a.
b.
c.
[A:] Carl proposed to Diane yesterday!
[B:] At last! She must have said ‘yes’.
[B:] ??At last! She said ‘yes’. [from Ninan (2014)]
(40-c) is marked in comparison with (40-b). Intuitively, what is going on in cases like this is that it is
permissible to propose the update in question on the basis of the group’s shared evidence, but it is not
permissible to propose it on the basis of one’s own assertoric authority. We can make sense of this if
we mildly strengthen the derivation of Support given in the last section, as follows. I argued there that
Support arises because pMust ϕq is felt to be in competition with the structurally simpler ϕ. Suppose we
strengthen this assumption so that we also predict that ϕ is felt to be in competition with pMust ϕq. In
other words, when a speaker asserts pMust ϕq or ϕ, interlocutors always ask why the speaker chose one
rather than the other. Then, since the relevant difference between these is that the first draws attention
to the conversants’ doxastic relation to JϕK, interlocutors will reason that the speaker will choose pMust
ϕq just in case she wants to update with JϕK on the basis of a shared argument, whereas she will choose
ϕ just in case she wants to update with JϕK on the basis of her own assertoric authority.
If we take this approach, then we can explain cases like (40) as follows. With (40-b), B asks A to
accept that Diane said ‘yes’ based on their shared evidence about Carl and Diane’s relationship. By
contrast, in (40-c), B asks A to accept that Diane said ‘yes’ on the basis of her own authority. In (40-c),
because B declines to choose the update option that references their shared information, she falsely
implicates that she has some independent source of evidence about what Diane said. In general, this
60 For
recent experimental evidence that speakers do indeed interpret pMust ϕq as in some sense weaker than ϕ alone, see Degen et al. (2016).
29
approach predicts that when some evidence in favor of JϕK is shared and we wish our interlocutors to
accept JϕK on the basis of that evidence, rather than our own authority, it will be obligatory to use a
‘must’-claim rather than a non-modal claim.
This proposal turns on a stronger assumption about alternatives than did our original derivation
(namely, that ϕ and pMust ϕq are always relevant alternatives), an assumption that needs to be further
explored and motivated. It is, however, a conservative extension of the account I have given, which
promises an interesting explanation of the felt weakness − and occasional obligatoriness − of ‘must’.
7
Conclusion
The main argument of this paper came in three stages. I began by arguing, on the basis of judgments
about cases and experimental data, that in addition to Indirectness, we need Support to fully characterize
the differences in felicity conditions between an assertion of pMust ϕq and an assertion of ϕ. Next, I
argued that Indirectness follows from Support plus general pragmatic principles about redundant assertions. Along the way, I sketched a new approach to pragmatic principles about redundancy. I argued that
this approach motivates a new representation of the common ground − and thus of propositional attitudes in general; and I argued that, given that approach, this derivation of Indirectness makes attractive
predictions about when a ‘must’-claim is unacceptable. Finally, I made a proposal about how to derive
Support from a modal semantics I have defended elsewhere. I then discussed two residual questions:
how to extend my account to embedded ‘must’, and to an explanation of the felt weakness − and, in
some cases, obligatoriness − of ‘must’.
The three main parts of this argument are, to a degree, independent: the second and third part of the
argument depend on the first part, but one could accept the first part without accepting the latter two, and
one could accept either of the latter two without accepting the other. If each of these moves is successful, however, then taken together, they constitute a solution to Karttunen’s Problem: characterizing and
explaining the differences in felicity conditions between an assertion of pMust ϕq and an assertion of ϕ:
because of its more complex form and the question it makes salient, pMust ϕq, unlike ϕ alone, requires
that an argument be given for JϕK; and from the requirement, in turn, we can conclude that the speaker’s
evidence for JϕK is indirect in a relevant sense.
In conclusion, I highlight three broad upshots of my approach to Karttunen’s Problem. The first is
about the relation between ‘must’ and evidentiality. I have argued that ‘must’ does not grammaticalize
a certain constraint on the type of evidence, in any intuitive sense, which the speaker must have for
its prejacent. Rather, the felt indirectness of ‘must’-claims is accounted for pragmatically, and it is
accounted for not directly in terms of judgments about type of evidence but rather in terms of judgments
about redundancy.
The second is about the meaning of ‘must’. My derivation of Support rests on a certain semantics
for ‘must’. If the derivation is successful, then it provides an argument for adopting that semantics, and
presents a challenge for advocates of different semantics for ‘must’: to show how those approaches can
explain Support.
30
The third regards the theory of redundancy. I have argued that the norm that governs redundancy
in assertions is stronger in some respects and weaker in others from the norm that has been assumed
in the literature. A theory of redundancy − in essence, a theory of how our minds structure and access
information − will play a central role in the philosophy of mind and cognitive science, and judgments
about discourses (with or without ‘must’) provide our richest source of data for these theories. I have
sketched a framework for thinking about redundancy, and a model of the underlying propositional attitudes. More work needs to be done to spell out the proposal I have made − in particular, what it amounts
to for a proposition to follow from an argument in a mutually obvious way − work which will allow us
to evaluate more precisely my claim that judgments about the indirectness of ‘must’-claims and about
the redundancy of assertions go together.
I close with an abstract point about the architecture of semantic and pragmatic theories. My proposal rests on the assumption that an assertion of pMust ϕq and an assertion of ϕ have the same basic
update effect, but different semantic values. Indeed, on the semantics I have sketched, pMust ϕq and ϕ
informationally entail one another, in the sense that a context updated with either one entails the other,
but they do not semantically entail one another. The possibility of this divergence between update effect
and meaning proved essential for simultaneously capturing the intuition that an update with pMust ϕq is
pragmatically strong, and the intuition that it is in some sense indirect. My proposal thus illustrates the
point that semantic content must be distinguished from pragmatic update effect in our theorizing about
natural language.61
References
Aikhenvald, A. (2004). Evidentiality. Oxford University Press, Oxford.
Barker, C. (2009). Clarity and the grammar of skepticism. Mind and Language, 24(3):253–73.
Bledin, J. and Rawlins, K. (2016). Epistemic resistance moves. Presentation at SALT 26.
Degen, J., Kao, J. T., Scontras, G., and Goodman, N. D. (2015). A cost- and information-based account of epistemic must. Poster at 28th
Annual CUNY Conference on Human Sentence Processing.
Degen, J., Scontras, G., Trotzke, A., and Wittenberg, E. (2016). Definitely, maybe: Approaching speaker commitment experimentally. MS,
Stanford University.
Faller, M. (2001). Remarks on evidential hierarchies. In Beaver, D., Kaufmann, S., Clark, B., and Casillas, L., editors, Proceedings of the
‘Semfest’. CSLI Publications: Stanford, CA.
Faller, M. (2012). Evidential scalar implicatures. Linguistics and Philosophy, 35:285–312.
von Fintel, K. and Gillies, A. (2010). Must...stay...strong! Natural Language Semantics, 18(4):351–383.
von Fintel, K. and Gillies, A. (2011). ‘Might’ made right. In Egan, A. and Weatherson, B., editors, Epistemic Modality. Oxford University
Press.
von Fintel, K. and Gillies, A. (2016). Still going strong: The semantics and pragmatics of epistemic must. Notes presented at 5th Annual OSU
Philosophy/Linguistics Workshop, http://mit.edu/fintel/kt-osu-notes.pdf.
Giannakidou, A. and Mari, A. (2016). Epistemic future and epistemic MUST: nonveridicality, evidence, and partial knowledge. In Blaszack,
J., Giannikidou, A., Klimek-Jankowska, D., and Mygdalski, K., editors, Mood, Aspect and Modality: What is a Linguistic Category?
University of Chicago Press.
Gibson, E., Piantadosi, S., and Fedorenko, K. (2011). Using Mechanical Turk to obtain and analyze English acceptability judgments. Language
and Linguistics Compass, 5(8):509–524.
Ginzburg, J. (1995a). Resolving questions. i. Linguistics and Philosophy, 18(5):459–527.
Ginzburg, J. (1995b). Resolving questions. ii. Linguistics and Philosophy, 18(6):567–609.
Goodhue, D. (2016). Epistemic must is not evidential, it’s epistemic. In Proceedings of the North East Linguistic Society (NELS), volume 46.
Grice, P. (1989). Studies in the Way of Words. Harvard.
Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. PhD thesis, University of Massachusetts, Amherst.
Heim, I. (1983). On the projection problem for presuppositions. In Barlow, M., Flickinger, D. P., and Wiegand, N., editors, The Second West
Coast Conference on Formal Linguistics (WCCFL), volume 2, pages 114–125. Stanford, Stanford University Press.
61 This
point is fairly uncontroversial, though inconsistent with certain interpretations of dynamic semantics.
31
Heim, I. (1992). Presupposition projection and the semantics of attitude verbs. Journal of Semantics, 9(3):183–221.
Ippolito, M. (2016). Constraints on the embeddability of epistemic modals. Handout at TOM 9 at McGill.
Karttunen, L. (1972). Possible and must. In Kimball, J., editor, Syntax and Semantics, volume 1, pages 1–20. Academic Press, New York.
Karttunen, L. (1974). Presuppositions and linguistic context. Theoretical Linguistics, 1(1-3):181–194.
Katzir, R. (2007). Structurally-defined alternatives. Linguistics and Philosophy, 30:669–690.
Katzir, R. and Singh, R. (2013). Hurford disjunctions: embedded exhaustification and structural economy. In Etxeberria, U., Fǎlǎuş, A.,
Irurtzun, A., and Leferman, B., editors, Proceedings of Sinn und Bedeutung 18.
Kratzer, A. (1981). The notional category of modality. In Eikmeyer, H. and Rieser, H., editors, Words, Worlds, and Contexts: New Approaches
in Word Semantics. de Gruyter.
Kratzer, A. (1991). Modality. In von Stechow, A. and Wunderlich, D., editors, Semantics: An International Handbook of Contemporary
Research, pages 639–650. de Gruyter, Berlin.
Kratzer, A. (2012a). Modals and Conditionals. Oxford University Press, Oxford.
Kratzer, A. (2012b). What ‘must’ and ‘can’ must and can mean. In Modals and Conditionals, pages 1–20. Oxford University Press.
Kripke, S. A. (2009). Presupposition and anaphora: Remarks on the formulation of the projection problem. Linguistic Inquiry, 40(3):367–386.
van Kuppevelt, J. (1995). Discourse structure, topicality and questioning. Journal of Linguistics, 31(01):109–147.
Lassiter, D. (2016). Must, knowledge, and (in)directness. Natural Language Semantics.
Lewis, D. (1979). Scorekeeping in a language game. Journal of Philosophical Logic, 8(1):339–359.
Lewis, D. (1988). Relevant implication. Theoria, 54(3):161–74.
MacFarlane, J. (2004). In what sense (if any) is logic normative for thought? Prsented at the Central Division APA.
Mandelkern, M. (2016a). How to do things with modals. Manuscript.
Mandelkern, M. (2016b). Shiftiness and shiftlessness: Epistemic modals and attitude ascriptions. Manuscript.
Martin, S. (2016). Supplemental update. Semantics & Pragmatics.
Matthewson, L. (2015). Evidential restrictions on epistemic modals. In Alonso-Ovalle, L. and Menendez-Benito, P., editors, Epistemic
Indefinites. Oxford University Press, New York.
Mayr, C. and Romoli, J. (2016). A puzzle for theories of redundancy: Exhaustification, incrementality, and the notion of local context.
Semantics and Pragmatics.
Montague, R. (1970). Universal grammaar. Theoria, 36:373–398.
Moss, S. (2015). On the semantics and pragmatics of epistemic vocabulary. Semantics and Pragmatics, 8:1–81.
Murray, S. (2014). Varieties of update. Semantics and Pragmatics, 7(2):1–53.
Ninan, D. (2014). Taste predicates and the acquaintance inference. In Snider, T., D’Antonio, S., and Weigand, M., editors, Proceedings of
SALT 24, pages 290–309.
Potts, C. (2005). The Logic of Conventional Implicatures. Oxford University Press, Oxford.
Roberts, C. (1998/2012). Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics,
5:1–69.
Schlenker, P. (2008). Be articulate: a pragmatic theory of presupposition projection. Theoretical Linguistics, 34(3):157–212.
Schlenker, P. (2009). Local contexts. Semantics and Pragmatics, 2(3):1–78.
Scott, D. (1970). Advice on modal logic. In Lambert, K., editor, Philosophical Problems in Logic, pages 143–173. Dordrecht Reidel.
Sherman, B. (2016). Open questions and epistemic necessity. Manuscript.
Silk, A. (2014). Discourse contextualism: A framework for contextualist semantics and pragmatics. Monograph, University of Birmingham.
Stalnaker, R. (1970). Pragmatics. Synthese, 22:272–289.
Stalnaker, R. (1973). Presuppositions. Journal of Philosophical Logic, 2:447–457.
Stalnaker, R. (1974). Pragmatic presuppositions. In Munitz, M. K. and Unger, P., editors, Semantics and Philosophy, pages 197–213. New
York University Press, New York.
Stalnaker, R. (1978). Assertion. In Cole, P., editor, Syntax and semantics, volume 9, pages 315–322. New York: Academic Press.
Stalnaker, R. (2002). Common ground. Linguistics and Philosophy, 25:701–721.
Stalnaker, R. (2014). Context. Oxford University Press.
Stephenson, T. (2007a). A parallel account of epistemic modals and predicates of personal taste. In Puig-Waldmueller, editor, Proceedings of
Sinn und Bedeutung, volume 11, pages 583–97.
Stephenson, T. C. (2007b). Judge dependence, epistemic modals, and predicates of personal taste. Linguistics and Philosophy, 30(4):487–525.
Stone, M. (1994). The reference argument of epistemic must. In Proceedings of IWCS 1, pages 181–190.
Swanson, E. (2015). The application of constraint semantics to the language of subjective uncertainty. Journal of Philosophical Logic, pages
1–26. To appear in the Journal of Philosophical Logic.
Veltman, F. (1985). Logics for Conditionals. PhD thesis, University of Amsterdam.
Willett, T. (1988a). A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language, 12(1):51–97.
Willett, T. (1988b). A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language, 12(1):51–97.
Williamson, T. (2000). Knowledge and its Limits. Oxford University Press.
Yablo, S. (2014). Aboutness. Princeton University Press.
Yalcin, S. (2011). Nonfactualism about epistemic modality. In Egan, A. and Weatherson, B., editors, Epistemic Modality, pages 295–332.
Oxford University Press.
Yalcin, S. (2016a). Figure and ground in logical space. Philosophy and Phenomenological Research.
Yalcin, S. (2016b). Modalities of normality. In Charlow, N. and Chrisman, M., editors, Deontic Modality. Oxford University Press.
32