Modality, Weights, and

Inconsistent Premise Sets

Alex Silk
asilk@umich.edu
Draf of April io1i
May, Ought, and Must they are contemptible auxiliaries.
George Eliot

Introduction
In this paper I will discuss several linguistic phenomena that appear to motivate
introducing various complexities into the semantics for weak necessity modals like
ought andshould. First, weak necessity modals appear tobe sensitive tothe strengths
of various considerations or ideals. Suppose that your only goals are to go for a run
(now) and to watch a movie (now). You cant do both, but you want to go for a
run much more than you want to watch a movie. Call this case wiicu1iu co.is.
Consider (1) and (i).
(1) In view of your goals, you should go for a run.
(i) In view of your goals, you should watch a movie.
I take the following intuitions to be reasonably robust: (1) is true and (i) is false.
On the face of it, however, existing analyses of modals seem to be silent about how
to handle such cases.
Second, there is a growing consensus that weak necessity modals like ought
and should are (as the names suggests) weaker than strong necessity modals like

Tanks to Nate Charlow, Dan Greco, Irene Heim, Dan Singer, Bob Stalnaker, Eric Swanson, and
Rich Tomason for helpful discussion.

Mary Garth, in Middlemarch, Bk. i, Ch. 1. Slightly modifed from the original.
1
must and have to. (For the purposes of this paper I will not distinguish between
the meanings of ought and should, on the one hand, or must and have to, on the
other.) Must is thought to asymmetrically entail Should , as the data in () and
() suggest.
1
() a. I should help the poor. In fact, I must.
b. I must help the poor. -In fact, I should.
() a. I should help the poor, but I dont have to.
b. -I must help the poor, but its not as if I should.
While should appears to allowstrengthening to must, as in (a), the converse does
not seem to hold, as shown in (b). Similarly, (a) is consistent in a way that (b) is
not. But if should and must are just universal quantifers over sets of worlds, it is
unclear how to capture this dierence in strength.
Some have argued that the standard quantifcational semantics for modals can-
not capture these phenomena, or that, if it can, it must render the semantics for weak
necessity modals signifcantly more complex than that for strong necessity modals.
I will argue that this is a mistake. Te standard semantics can capture the relevant
data and can do so while minimizing the dierences between the lexical entries for
weak and strong necessity modals. Te resulting analysis, I hope, will clarify what
sorts of premises fgure in the interpretation of modals, and illuminate previously
puzzling and unnoticed semantic and pragmatic properties of weak and strong ne-
cessity modals.
e classical semantics
As our task is to examine whether the standard semantics can accommodate data
like that described in 1 it will be helpful to briefy characterize what this standard
semantics is. On the classical semantics for modals, as made well known in linguis-
tic semantics by Angelika Kratzer (1,,, 1,, 181a, 11), modals are treated as
context-dependent quantifers Qover possible worlds. Simplifying a bit, the domain
of quantifcation is set by two contextually supplied parameters: a set f of accessible
worlds (a modal base), and a preorder (a refexive and transitive relation) on
1
See, e.g., Siom. 1,o, Hov 1,i, 18, Wiv1uiimiv 1,i, ch. , H.vm. 1,,, 1,
Lvos 1,,, Wiiii.ms 181, Jois & Pov 18o, P.imiv ioo1, Coviiv iooo, vo Fi1ii & I.-
1viuoU ioo8.
i
V, where this preorder ranks worlds along some relevant dimension.
i
Roughly, a
modal claim says that the modals prejacent the proposition in its scope holds
in Q of the worlds in the modal base that rank highest in the preorder.
Dierent readings of modals arise from dierent contextual resolutions of the
modal base and preorder. Modal bases determine refexive accessibility relations:
they are sets of propositions describing a body of relevant circumstances or evidence
in the world of evaluation. Te preorders are determined by an ordering source
g(w) a set of propositions indexed to the world of evaluation in the following
way: for any worlds w

and w

, w

is at least as good as w

relative to the ideal set up

by g(w) i all propositions in g(w) that are true in w

are also true in w

.
Denition . w

g(w)
w

= p g(w)(w

p w

p)
An ordering source describes the content of some ideal (e.g., the law, morality, your
goals, the usual course of events). For example, the phrase In view of your goals
calls for an ordering source consisting of the set of propositions describing your
goals in the world of evaluation.
Given a preordered set (V,
g(w)
), a selection function Dcan be defned to select
from some domain those worlds that are not
g(w)
bettered by any other world:
Denition . Z V D(Z,
g(w)
) = {w

Z w

Z w

g(w)
w

g(w)
w

}
D selects the set of
g(w)
-maximal (
g(w)
best) worlds from the modal base, those
worlds in the modal base that best approximate the relevant ideal. Modals quantify
over these worlds in D(
f(w),
g(w)
). Te truth conditions for Should , on the
standard semantics, are as follows. (Ill address relevant dierences between weak
and strong necessity modals shortly.)

Denition . Should
c, w
= 1 i w

D(
f(w),
g(w)
)
c,w

= 1
Tis says that Should is true i is true at all the
g(w)
best accessible worlds.
i
See also v. Fv..ssi 1,, Liwis 1,, Vii1m. 1,o. Kratzers 1,,, 1, analyses are
couched in a premise semantic framework; but given the equivalence of premise semantics and or-
dering semantics as described in Liwis 181 no harm results from my couching the discussion in
terms of ordering frames.

I sometimes use modal base to refer to a set of propositions, sometimes to the intersection of
this set; I trust that context will disambiguate. And I make the Limit Assumption and assume that
our selection function is well-defned and non-empty (see Liwis 1,, 181, Kv.1ziv 181a, 11,
Sw.so io11 for relevant complications.)

Weights and priorities

Cases like wiicu1iu co.is pose a prima facie problem for the standard semantics.
Given the intuitive characterization of ordering sources above, wiicu1iu co.is
would seem to call for an ordering source like in (,) that describes the contents
of your goals. (Hereafer I will leave implicit that the goals in question are for the
present time.)
(,) g(w) = {You go for a run, You watch a movie}
Simplifying a bit, suppose the (circumstantial) modal base consists of the worlds
(or representatives of relevant equivalence classes of worlds) R and M, where you
go for a run in R and watch a movie in M. Since you cannot simultaneously go
for a run and watch a movie, there is no world in the modal base in which both
of your goals are satisfed. Tus, by Defnition 1, R and M are
g(w)
incomparable:
R {You go for a run} and M {You watch a movie}, but neither {You go for a run}
nor {You watch a movie} is a subset of the other. By Defnition , this seems to
incorrectly predict that both (1) and (i) are false.

R and M are both

g(w)
best, and
its not the case that inbothof themyougofor a runor that inbothof themyouwatch
a movie. Intuitively, however, the dierential strength of your goals should be able to
break the tie between Rand M. Evenif you dont have to go for a runand dont have
towatcha movie, (1) shouldbe true and(i) shouldbe false. Te problemis that there
seems to be now way of representing the dierent strengths of your goals. Should
sentences are evaluated with respect to maximally consistent subsets of g(w) and
there is nothing that distinguishes {You go for a run} from {You watch a movie}.
Some have taken this problemto be insurmountable for the standard semantics.
For example, here is Lassiter: Te problemis fundamentally that [Kratzers] theory
makes no room for one [goals] being stronger than another; instead any confict of

On an adaptation of Swansons io11 semantics for ought that makes the Limit Assumption,
ought is distinguished from must in that the former quantifes over the maximal elements of
some
g(w)
chain, whereas the latter quantifes over the maximal elements of every
g(w)
chain (cf.
v. Fv..ssi 1,). On this semantics, it is must that is given the truth conditions in Defnition ,
and ought is restricted to quantifying over the worlds in D(f(w) S,
g(w)
), for some maximal

g(w)
chain S. (A chain is a totally preordered subset of a partial preorder of a set of worlds; a
maximal chain is a chain that is not a proper subset of any other chain.) I do not consider this
complication in this section since it will not by itself help with our current problem. It renders both
(1) and (i) true, whereas what we need is some way of distinguishing their truth values. However,
the remainder of my discussion is consistent with these Swanson-inspired truth-conditions. (See
Siix io1i for discussion of how this semantics may be used to help capture certain dierences in
information-sensitivity between weak and strong necessity modals.)

[goals] leads to incomparability; the theory doesnt leave any roomfor [goals] be-
ing stronger or weaker than another[goals] are all-or-nothing (io11, pp. o1oi,
oo; adapted for the case of teleological modals, as opposed to epistemic modals).
I will argue that this problem is spurious.
One initially plausible way of revising the standard semantics to capture prior-
ities might involve introducing a weighting relation
w
that ranks the propositions
in the ordering source. (Suppose
w
is also a (possibly partial) preorder, here on a
set of propositions.) We could then introduce a function F
w
that takes
w
as argu-
ment as returns a preorder on the power set of this set. Tis would yield a preorder
on the maximally consistent subsets of g(w). To a very rough frst approximation
(and making an analogue of the Limit Assumption), we could then say that should
quantifes over the best worlds in the F
w
(
w
) best maximally consistent subset of
g(w). Inthe case of wiicu1iuco.is, F
w
andwouldlet us privilege the maximally
consistent subset {You go for a run} in the interpretation of should; the ordering
on worlds would be determined only by this set of propositions rather than by the
set given in (,).
Perhaps suchcomplications andsurely more, were this strategy pursued may
be needed to capture priorities or weights in the interpretation of should. But there
are good reasons for being optimistic that they are not. Intuitively it is clear what the
preorder called for in wiicu1iu co.is should look like: R should be ranked higher
than M. Given the equivalence of ordering semantics and premise semantics (Liwis
181), we can be assured that there is some set of propositions that induces this or-
dering: We must be selective in the choice of premises By judicious selection,
we can accomplish the same sort of discrimination as would result from unequal
treatment of premises (Liwis 181, pp. iioii1; cf. Kv.1ziv 181b, p. i1o).
,
Te
question, of course, is whether we can fnd a way of determining what this set of
propositions is, given the contents of your goals, that is not ad hoc.
o
,
Tanks to Kai von Fintel for reminding me of the importance of this point.
o
It might be thought that lumping, in the sense of Kv.1ziv 181b, 18, iooi, could help
represent weights and priorities in cases like wiicu1iu co.is. Discussing the case of the relative
importance of various facts in the interpretation of counterfactuals, Kratzer argues that we can avoid
introducing priorities by lumping certain propositions together. By lumping a proposition into
more premises, it becomes harder to give up; this refects its relative priority. But it is unclear how
lumping can help us. For one thing, on Kratzers ocial defnition of lumping, roughly, one proposi-
tion lumps another in a world i any part of the world (situation) that makes the one true makes the
other true (18, p. o1o). One of the main ideas of Kv.1ziv iooi is that at least for the interpre-
tation of counterfactuals whenever a premise set contains propositions p and q such that p lumps
q, p also entails q. Given these conditions for lumping, it is hard to see how either proposition in (,)
lumps the other. Perhaps an alternative way of understanding lumping can yield a more successful
,
Te problem, I suggest, is that we do not have a nuanced enough characteriza-
tion of the contents of your goals in wiicu1iu co.is. Typically our goals are not to
do such-and-such no matter what. Norms, desires, goals, and the like do not usually
come in form of blunt Kantian categorical imperatives or divine commands in
the case of wiicu1iu co.is, something like Go for a run, no matter what! or Vatch
a movie, no matter what!. Rather they ofen come with (what I will call) conditions
of applicability; they can be hypothetical, or prima facie (cf. Ross 1o, H.vi 1,i,
K.1 1,o). For instance, in the moral case, a premise that describes a prohibition
against lying need not be categorical, something like You never lie. Instead it might
be conditional on the absence of certain defeating conditions, e.g., something like
Unless not lying would lead to an innocent persons death or some other serious harm
or, you do not lie. Simplifying quite a bit, in wiicu1iu co.is the ordering source
might be more accurately represented by something like in (o), or, more schemati-
cally, like in (,), where C describes the relevant applicability conditions, the condi-
tions under which (and, in this case, only under which) the goal in question is to be
pursued. (For simplicity, the conditionals can be understood as expressing material
implication.)
(o) g(w) =
{ You go for a run i you dont want to do anything else more,
You watch a movie i you dont want to do anything else more }
(,) g(w) =
{ You go for a run i C,
You watch a movie i C

}
In this way, rather than introducing a semantic mechanism that weights categorical
propositions in the ordering source, we can capture these weights by encoding rele-
vant features of the priorities inquestioninto the premises themselves specifcally,
into the norms applicability conditions. (Unconditional premises can be treated as
having V as their applicability condition.)
A simplifed ordering source like the one in (o) generates the correct results in
wiicu1iu co.is. Given that one of the relevant circumstances is that you want to
go for a run more than you want to watch a movie, Cbut not C

will hold in R and M,

the worlds in the circumstantial modal base. Accordingly R will be
g(w)
better than
M, by Defnition 1: R verifes both propositions in the ordering source (vacuously
in the case of the second), whereas M verifes neither proposition. So we correctly
analysis. But since I will argue that we can avoid introducing priorities by giving a more nuanced
characterization of the contents of the relevant goals, I will not pursue the lumping strategy here.
o
predict that (1) is true and (i) is false.
In this way, though all of the propositions in the ordering source fgure into the
interpretation of the modal, some of them may do so only trivially. If their applica-
bility conditions are not met, they may be vacuously satisfed throughout the modal
base; if so, they will not distinguish among these worlds. Tis captures the intuitive
idea that given the relevant circumstances that is, given which premises have their
applicability conditions met only certain of the norms, goals, etc. are in force.
It remains to be seen, of course, whether this strategy will generalize to more
complex cases. But insofar as we have clear intuitions about what preorders are
called for, I see no reason why more nuanced characterizations of the applicability
conditions for the norms, goals, etc. in question will not yield the correct results.
To be sure, in ordinary conversations it may be underdetermined precisely what
these applicability conditions are. But this is just one more way in which modal
expressions can be vague.
Ought and must, most and all, best and only
Intuitively should seems to be used to express that an action or proposition is nec-
essary. Yet it is well established that though should is stronger than possibility
modals like may or can, it is nevertheless weaker than modals like must and have
to. What is not well established is how to capture this dierence in strength. In the
remainder of this paper I will argue that applicability conditions play a crucial role
in capturing certain aspects of the weak/strong necessity distinction.
As noted in 1, must appears to be logically stronger than should. Must sen-
tences seem to asymmetrically entail their counterparts with should.
() a. I should help the poor. In fact, I must.
b. I must help the poor. -In fact, I should.
() a. I should help the poor, but I dont have to.
b. -I must help the poor, but its not as if I should.
Tis behavior appears to parallel familiar dierences between nominal quantifers,
as refected in (8) and ().
(8) a. Some (/most) of the students got an A. In fact, all of them did.
b. All of the students got an A. -In fact, some (/most) of them did.
() a. Some (/most) of the students got an A, but not all of them.
,
b. -All of the students got an A, but not some (/most) of them.
In reply to this data one might treat weak necessity modals as most-quantifers
over the accessible worlds.
,
On this view, whereas strong necessity modals quantify
over all of the best worlds, weak necessity modals only quantify over most of them.
However, as vo Fi1ii & I.1viuoU (ioo8, p. 118) note, our semantics should not
presuppose that there are fnitely many worlds or that we cancount them(cf. Coviiv
iooo).
Instead vo Fi1ii & I.1viuoU (ioo,, ioo8) take up the plausible idea from
Siom.1,o that weak necessity modals quantify over the best of the best worlds,
or over the
g(w)
best worlds that are also best by some additional measure. Whereas
weak necessity modals [pick] out the best means without excluding the possibility
of others, strong necessity modals [imply] that no other means exist (Siom.
1,o, pp. o1). As Bernard Williams puts it, Ought is related to must as best is
related to only (181, p. 1i,). For example, (1o), a now familiar example, conveys
that there multiple ways of getting to Harlem but that taking the A train is best by
some additional measure.
(1o) To go to Harlem, you should take the A train.
Saying that you should take the A train to go to Harlem because its quickest is, on
this view, the same as saying that you must take the A train to go to Harlem in the
quickest way (cf. vo Fi1ii & I.1viuoU ioo8, pp. 11811, 1,).
vonFintel and Iatridou implement this idea by introducing a secondary ordering
source g

(w) to refne the initial preorder

g(w)
. Whereas strong necessity modals
(universally) quantify over the
g(w)
best worlds, weak necessity modals (univer-
sally) quantify over a subset of these worlds. Tey quantify over the
g(w)
best worlds
that are also best in view of g

(w) to a frst approximation, over the worlds in

D(D(
f(w),
g(w)
) ,
g

(w)
). (See their ioo8, p. 18n., for potential refnements.)
In the case of (1o), the primary ordering source might be something like {You go to
Harlem} whereas the secondary ordering source might be something like {You take
the quickest route when going to Harlem}.
Tough this analysis captures how should is weaker than must while avoiding
counting worlds, there is something unsatisfying about it. Importantly, little is said
about what makes a primary ordering source primary and a secondary ordering
,
Coviiv iooo attributes this idea to Hov1,i, 18. However, it is not clear that Hornactually
defends it. He just seems to argue that ought behaves in similar ways to most e.g., they both allow
neg-raising which suggests that they both are mid-range elements on their respective scales.
8
source secondary apart from the fact that the latter fgures into the interpretation
of weak but not strong necessity modals. No story is given about how primary and
secondary ordering sources are determined independently of the truth conditions of
the relevant should and must sentences, that is, in a manner that does not simply
involve reverse engineering them from relevant truth value judgments.
Ought, must, and the common ground
In recent work, Aynat Rubinstein (io11, io1ia, io1ib), as I understand her, at-
tempts to capture the intuitions driving the analysis invoFi1ii &I.1viuoUioo8
without appealing to primary and secondary ordering sources. According to Rubin-
stein, the apparent distinction between primary and secondary ordering sources is
in fact a distinction between considerations that are and are not currently accepted
in the discourse common ground (the set of live possibilities for the purposes of the
conversation; S1.i.xiv 1,, 1,8) or the conversational scoreboard (Liwis
1,b). Te apparently primary ordering source consists of precisely those propo-
sitions describing the goals, norms, etc. that are currently accepted in the common
ground. Te apparently secondary ordering source that weak necessity modals
are sensitive to consists of assumptions that are not currently common ground. On
Rubinsteins view there is in fact only one kind of ordering source. However, weak
necessity modals conventionally signal that the truth of the necessity claimrelies on
assumptions that outstrip the common ground. By contrast, for a strong necessity
modal to be felicitous its truth cannot depend on considerations not accepted in the
common ground.
8
Strong necessity modals quantify only over the best worlds as
determined by the conversationally agreed upon values of f(w) and g(w), whereas
weak necessity modals conventionally signal that for the prejacent to be necessary
in the relevant sense the modal must be interpreted with respect to a monotonic
update of f(w) or g(w), that is, a superset of f(w) of g(w).
Rubinstein proposes the following truth conditions and associated conventional
signal, SigW, adapted in light of our conventions (io11, p. 1):
8
Not accepted when: I would assume that it is afer the statement has been made, as S1.i.xiv
iooi has taught us. Even if the necessity of relies on assumptions beyond the discourse common
ground as it was prior to the utterance, Must may still be felicitous because, as per the performative
aspect of the meaning of must, a successful utterance of Must changes the context and the
relevant parameters of modal base and preorder to make it such that it entails the necessity of
. On the performativity of must see especially Ni. ioo,, Pov1iv ioo,, pp. oo,, ioo,
pp. 1o1o,, 181o; cf., e.g., Lvos 1,,, Liwis 1,a.

(11) Should
c, w
= 1 i w

D(
f
+
(w),
g
+
(w)
)
c,w

= 1, where it is con-
ventionally signaled that f
+
(w) and g
+
(w) are (possibly trivial) monotonic
updates of the conversationally agreed upon parameters f
cg
(w) and g
cg
(w).
She argues that this signal is a type of conventional meaning that is projective, not-
at-issue, antibackgrounding in a sense, and grammatically determined. Tus, it is
most similar to a type of presupposition that is nonetheless not pre-supposed or
backgrounded in any intuitive sense (io11, p. 11). (We will return to this point
below.) By supplementing the lexical entry for weak necessity modals with this
(non-truth-conditional) signal, Rubinstein can avoid complicating the denotation
for should in the manner of vo Fi1ii & I.1viuoU ioo8.
Rubinsteins account nicely highlights previously underappreciated sensitivities
to common ground assumptions in weak and strong necessity modals. For example,
suppose I want to go to Harlemand amconsidering howto get there. What has been
observed in the literature is that if (it is common ground that) there is only one way
to get to Harlem, it is the strong but not the weak necessity modal that is felicitous.
(1i) You must (-should) take the A train.
But what has not been adequately appreciated is that in certain contexts it can still
be must that is preferred even if (it is common ground that) the A train is not the
only option. Suppose I amconsidering what route to take, and it is common ground
that there are several available options. I mention that my traveling quickly is im-
portant to me, and you know that the A train is the quickest route. As Rubinsteins
account predicts, your communicating that my taking the A train is teleologically
necessary depends on how the status of my preference for speed and certain other
of my goals fgure in the common ground (cf. RUvis1ii io11, p. 11). (I will use
a ? superscript to mark that using the item is dispreferred, and a - to mark that
using the item is infelicitous; thus - marks the more deviant usage.)
(1) [Context: I want to go to Harlem. Tere has been no prior mention of my
other preferences.]
a. Me: Traveling quickly would be nice.
You: Sounds good. You should (
?
must) take the A train.
b. Me: Traveling quickly is most important more important thansafety
or anything else.
You: Sounds good. You must (
?
should) take the A train.
Te necessity of my taking the A train depends on my preference for speeds being
1o
more important in this context than other potentially competing preferences. Te
weak necessity should is felicitous only if the status of this preference for speed
vis--vis other potentially relevant preferences is not currently established in the
conversation, as in (1a). According to Rubinstein, by using should you conven-
tionally signal that the teleological necessity of your taking the A train follows from
this non-common-ground assumption. But once it becomes common ground that
my preference for traveling quickly is to take precedence over any potentially com-
peting preferences, it is must, not should, that is preferred, as in (1b). In this
way what counts as the only means is essentially context-dependent, dependent
on what is currently accepted in the discourse common ground.
Rubinsteins account opens up exciting new theoretical possibilities for distin-
guishing weak and strong necessity modals. Te relative felicity of weak and strong
necessity modals does seem to interestingly depend on properties of the discourse
common ground. However, I have several worries with the details of her proposal.
My aimin what follows is not to provide a knockdown case against her account or to
argue that her account cannot be made to work. Rather my aimis to give us reasons
to look for an alternative. Working through these concerns will be helpful in this
regard.
Rubinstein argues that the signal SigW in (11) is a conventional signal, a non-
truth-conditional aspect of the conventional meaning of weak necessity modals.
But it is not clear whether SigW passes the tests she considers, or whether there
is independent evidence for an expressions having the kind of meaning component
proposed for SigW. First, Rubinstein argues that SigW is not-at-issue content on
the ground that it passes the indirect denial version of the Hey, wait a minute.
When an utterance generally carries the not-at-issue content that , cannot be
directly denied by (e.g.) No!, but only indirectly by (e.g.) Hey, wait a minute! or
Whaddaya mean!, as in (1).
(1) [Context: It is not common ground that Alice has smoked in the past.]
Me: Alice stopped smoking.
You: (-Tats not true!) Hey, wait a minute! Alice didnt use to smoke!
Rubinsteinclaims that SigWcanonly be denied indirectly inthis way, and that direct
denial can only target the quantifcational element of the meaning of should (cf.
io11, p. 1). (Here and in what follows the examples are mine though they are
intended to be parallel to the ones she gives.)
(1,) [Context: It is commonground that we want to go to Harlemand that speed
11
of travel is most important in deciding among routes.]
You: We should take the A train.
a. Me: Tats not true!
?
Traveling quickly is our stated goal.
b. Me: Whaddaya mean should: Traveling quickly is our stated goal;
we must take the A train.
I agree with Rubinstein that your using should in (1,) is dispreferred, but I
disagree with her diagnosis. To see why, note that, surprisingly, though SigW is
claimed to be not-at-issue, it is also claimed not to be backgrounded. Tough SigW
passes the indirect denial version of the Hey, wait a minute test, Rubinstein claims
that it fails the version of the test from Su.o 1,o and vo Fi1ii ioo (io11,
p. 1o). On this version of the test, the hearer not only interjects but also echoes the
content that that was not previously established in the common ground, as with a
follow-up like I had no idea that .
(1o) [Context: It is not common ground that Alice has smoked in the past.]
Me: Alice stopped smoking.
You: Hey, wait a minute! I had no idea that Alice used to smoke.
Rubinstein claims that such I had no idea that follow-ups are infelicitous follow-
ing a weak necessity claim. (I quote her gloss of SigW directly.)
(1,) [Context: It is common ground that we want to go to Harlem. It is not
common ground whether there are multiple ways for us to get there, or
what criteria would be used in deciding among routes.]
You: We should take the A train.
Me: Hey, wait a minute!
?
I had no idea you were making additional as-
sumptions. (/
?
I had no idea you were taking into account additional
preferences.)
She takes dialogues like in (1,) as evidence that SigW is not pre-supposed or back-
grounded information In fact, SigW has a special kind of antibackgrounding ef-
fect (io11, p. 1o). Using should highlights, in a certain sense, that non-common-
ground assumptions are the basis of the speakers necessity claim, and it does so in
such a way that makes natural a request from the addressee that the speaker make
these assumptions public.
I have several worries here. First, it seems that my response in (1,) would be
just as appropriate had you used must instead of should. Tis gives us reason to
think that it is not a special conventional property unique to weak necessity modals
1i
that is being targeted by my response. More pressingly, that SigW should jointly
have the alleged pair of properties being not-at-issue and antibackgrounded is
puzzling. Typically, expressions pass both versions or neither version of the Hey,
wait a minute test. It is not obvious what it means to say that a kind of content is
simultaneously not-at-issue and yet antibackgrounded, and I am not aware of any
(other) expressions that have contents with both of these properties. Tough Ru-
binstein suggests that SigW is a kind of Class D projective meaning in the sense of
Bi.viv i1 .i. ioob, meanings of this kind are conventionally backgrounded impli-
cations (like the prejacent of only and the polar component of approximatives). In
fact Beaver et al. gloss non-at-issue as backgrounded (ioob, pp. 1o11). Tis
gives us reason to think something is awry in Rubinsteins account.
Second, Rubinsteinclaims that SigWis a kindof non-cancelable projective mean-
ing onthe groundthat it passes the so-called Family of Sentences tests (Cuiivcui.
& McCoiii-Gii1 1o; cf. K.v11Ui 1,, Bi.viv ioo1, Bi.viv i1 .i.
iooa). SigW is alleged to survive as an utterance implication even when the weak
necessity modal is embedded in an entailment-canceling environment, as in a ques-
tion, in the antecedent of a conditional, or under negation (io11, pp. 1,1). For
purposes of space I will not go through each of these cases in depth. My only pur-
pose is to raise doubts concerning Rubinsteins data and whether SigW is necessary
to explain it.
Te following examples summarize the data Rubinstein cites. (I use the strong
necessity have to in some of these examples since they occur more naturally than
must in the relevant environments.)
(18) [Context: My friend needs to get to Harlem, and I am trying to determine
which train she should take. Te ticket booth worker says that both the
A train and the C train will get her there. I realize my friend didnt say
anything about how soon she needs to get to her destination.]
Me: Should she (/
?
Does she have to) take the A train:
Worker: Yes, if she wants the faster option.
(1) Worker: Te A train is faster, but it doesnt have any room for bikes.
Me: If (in light of her schedule) she should take the faster train, shell
have to leave her bike at home.
(io) [Context: Our groups goal is to get to a concert in Harlem. We dont all
have to take the same route. Te Atrain is quicker, and the Ctrain is safer.]
Me: Its possible for us to get to Harlem without us all taking the A train,
but the chances that wed all be able to sit together would be less. To
1
be confdent well all be able to sit together, all of us should take the
A train.
You: Some of us would rather be safe than sorry, so its not the case that all
of us should take the A train.
Start with (18). First, Rubinstein claims that substituting the strong necessity have
to for the weak necessity should in my question is infelicitous, given that we have
only mentioned my friends goal of getting to Harlem. I disagree with Rubinstein
here; I fnd the variant with have to perfectly acceptable. Intuitively, I amasking the
worker whether there are further (reasonable) considerations that make taking the
A train uniquely necessary. Assuming that have to is felicitous, we have reason to
think that it isnt SigW that explains why should is felicitous in (18). Second, SigW
seems to predict that it should be very hard to deny a question with a weak necessity
modal. Especially in contexts where little is settled in the common ground, there
will typically be some (monotonic) update of the commonground modal parameters
with respect to which the prejacent is necessary.
Second, turn to the conditional in (1). Rubinstein acknowledges that the felic-
ity of should in the if -clause need not be taken to be evidence for projection of
SigW. Since (a) the if -clause itself aects contextually supplied modal parameters,
and (b) the strong necessity have to can be substituted for should without infe-
licity, she grants that it may be the conditional and not SigW that is responsible for
the apparent projection of the inference concerning the speakers grounds for her
utterance.
Tird, consider the case of negation. Again, Rubinstein notes that have to can
felicitously replace should in (io). But since this is simply because have to is log-
ically stronger than should, the point that weak necessity modals carry a certain
conventional signal is not necessary to explain the data. An account on which weak
and strong necessity modals make dierent contributions to truth conditions would
presumably be able to capture the point just as well perhaps better, as well see;
but Im getting ahead of myself.
For these reasons it is unclear that the behavior of should in the Family of Sen-
tences tests provides evidence for projection of SigW. In each case it does not seem
that weak and strong necessity modals appear in complementary distribution, as
Rubinsteins account might seem to predict in light of its claim that they carry in-
compatible non-cancelable conventional signals. Of course this is not to say that
the data is inconsistent with SigW. My only point at the moment is that the data
presented thus far does not seem to uniquely favor an account according to which
weak necessity modals, unlike their strong counterparts, carry a projective, non-
1
truth-conditional component to their conventional meaning.
Finally, Rubinstein notes that an utterance of Ought does give rise to one kind
of cancelable implication, namely, the upper-bounding scalar implicature that Must
is false (or at least that the speaker does not have sucient evidence that Must
is true), as we saw in ().
() a. I should help the poor. In fact, I must.
b. I must help the poor. -In fact, I should.
However, it is unclear whether an account like Rubinsteins is best situated to cap-
ture how (a) is felicitous. Tis is because on Rubinsteins account weak and strong
necessity modals have the same truth conditions and dier only in the possible pres-
ence of a conventional signal concerning the modals contextual parameters. Since,
for Rubinstein, should signals that the truth of the necessity claimrelies on assump-
tions that outstrip the common ground, if the follow-up must sentence in (a) is in
fact true and felicitous, it would seem that the preceding should sentence should
be anomalous. But it is not. Again, if weak and strong necessity modals carry in-
compatible non-cancelable conventional signals, we would not expect both of them
to be able to appear felicitously in the same context. But they can.
Tis highlights one fnal concern with Rubinsteins account. Since weak and
strong necessity modals are claimed to make the same contribution to truth condi-
tions, one might worry that the proposed conventional signal amounts to something
of a stipulation. Since conventions are, in a sense, essentially arbitrary, an account
which derived a signal like SigW from (say) the truth conditions of weak necessity
modals and more general, independently attested conversational principles may be
preferable. Te positive account in o will attempt to do just that.
To recap, I want to emphasize that my goal in this section has not been to argue
that Rubinsteins account cannot be made to work. My aim has been to raise several
concerns with accounting for the dierence in weak and strong necessity modals in
terms of a projective, non-cancelable conventional signal of the sort that Rubinstein
describes. I have raised worries about whether the tests she oers provide genuine
evidence for SigWindeed, that weak and strong necessity modals both appear
to be felicitous in certain of her examples may even provide counterevidence. Te
behavior of weak necessity modals in familiar upper-bounding scalar implicatures
and when embedded under negation give us further reason to reconsider the idea
from that the (or a) main dierence between weak and strong necessity modals
is a truth conditional dierence in quantifcational strength. In the next section
I will defend this idea and argue that we can still retain Rubinsteins core insights
1,
concerning the weak/strong necessity distinction and the common ground.
Ought, must, and counterfactual assumptions
In order to motivate my positive proposal, lets return to the examples in (1) and
clarify precisely what the conversationally agreed upon values of modal base and
ordering source are and what (non-)common-ground assumptions are in play. One
initially plausible thought is that in (1a) the extra assumption concerns the goal of
traveling quickly that fgures in the ordering source. Tough Rubinstein does not
say explicitly what the precise values of f
cg
(w) and g
cg
(w) are (though perhaps she
does in other work), she seems to have this sort of idea in mind:
Should is felicitous in case the goal of traveling quickly is not yet estab-
lished in the discourse (1a). Te minute it is made clear that this more
detailed goal is the one under discussion, as in (1b), must becomes ap-
propriate and should no longer is In (1), the additional assumption
is that I will travel quickly. It enters the set of propositions that de-
scribes the contextually determined priorities or preferences, i.e., the
ordering source. (RUvis1ii io11, p. 11; adapted for our example)
Saying that the goal of traveling quickly is not yet established in the discourse in
(1a) and that it fgures into the ordering source might suggest that g
cg
(w) = {I go
to Harlem} and g
+
(w) = {I go to Harlem, I travel quickly} (cf. io11, p. 1o).
But this doesnt seemright. It is unclear why the goal of traveling quickly should
not be in g
cg
(w) afer my utterance in (1a). Afer all, I have expressed my prefer-
ence and you have acknowledged it. It is accepted that the frst-order goal of travel-
ing quickly is to be promoted. Rather it seems that the extra assumption concerns
whether there might be competing goals and, if there are, how to rank their relative
importance. As our discussion in suggests, my utterance in (1a) does establish
in the common ground my goal of traveling quickly, but it does so only condition-
ally. What gets accepted into the common ground, in both (1a) and (1b), is that
my goal of traveling quickly is to be promoted given that certain defeating conditions
do not obtain. What is not settled in (1a), but which is settled in (1b), is whether
these defeating conditions do obtain. In both (1a) and (1b) the conversationally
agreed upon value of the ordering source may be the same perhaps something
like the following:
1o
(i1) g
cg
(w) = {I go to Harlem, I travel quickly i I dont have any relevant stronger
countervailing goals}
However, whereas the proposition that I do not have any relevant stronger counter-
vailing goals the applicability conditionCfor my goal of traveling quickly holds
throughout f
cg
(w) in (1b), it does not in (1a). Since the teleological necessity of
my taking the A train depends on the assumption that this applicability condition
C holds, we predict that should should be preferred in (1a) and must should be
preferred in (1b). (If the modal base is not restricted to C-worlds, there will be
C-worlds in which I go to Harlem but dont take the A train that are
g
cg
(w)
best;
they will verify both propositions in g
cg
(w), vacuously in the case of the second.)
Tis analysis might seem to be congruent with the spirit (if not some of the de-
tails) of Rubinsteins account. But it suggests something surprising that, I take it,
gives us reasonto consider analternative. What is surprising is that eventhough(1)
concerns teleological modals, the modals in (1a) and (1b) may be interpreted with
respect to the same ordering source; the possibly non-common-ground assumption
is one that fgures into the modal base. I suggest that herein lies the important dier-
ence between weak and strong necessity modals and their relation to the common
ground. Roughly, the proposal is that strong necessity modals presuppose that the
applicability conditions of certain relevant norms, goals, etc. hold, whereas weak
necessity modals make a subjunctive claim about what would be necessary in the
relevant sense were those applicability conditions to hold.
More formally, for an ordering source g(w), consider the set {C
1
, C
i
, C

, . . . , C
n
}
of applicability conditions for each of the premises in g(w). And consider the set
{C
1
, C
i
, C

, . . . , C
n
} whose members are, for each C
k
, the complement of C
k
. Let
C
g(w)
be a set constructed by taking, for each k, either C
k
or C
k
. To a frst approxi-
mation, I suggest the following presuppositions and truth conditions for weak and
strong necessity modals:

.c1U.i s1voc
Presupposition:
For modal base f
cg
(w), ordering source g
cg
(w), and set of applicability condi-
tions C
g
cg
(w)
f
cg
(w) C
g
cg
(w)
Truth conditions:
Must
c, w
= 1 i w

D(f
cg
(w),
g
cg
(w)
)
c,w

= 1

Tough I treat the claim concerning the relation between the modal base and the set of ap-
plicability conditions to be a presupposition as opposed to an implicature or part of the truth
conditions I do not take much of the substance of what follows to hinge on this.
1,
sUv,Uc1ivi wi.x
Presupposition:
For modal base f
cg
(w), ordering source g
cg
(w), and set of applicability condi-
tions C
g
cg
(w)
(f
cg
(w) C
g
cg
(w)
)
Truth conditions:
Should
c, w
= 1 i w

D((f
cg
(w) C
g
cg
(w)
),
g
cg
(w)
)
c,w

= 1
First, it is easy to see that these truth conditions capture how must is logically
stronger than should, that is, that Must asymmetrically entails Should . In any
context in which the presupposition of must is satisfed, the set of worlds must
universally quantifes over is a superset of the set of worlds should quantifes over,
by Defnition i: must quantifes over the maximal elements of
g
cg
(w)
(
f
cg
(w))
i
,
whereas should quantifes over the maximal elements of the order preserving sub-
preorder
g
cg
(w)
((f
cg
(w) C
g
cg
(w)
))
i
. Tus, if Must is true, Should must
be true, but not vice versa. (Compare the asymmetric entailment between Every
child ate a cookie and Every child with a green shirt ate a cookie.)
Tese entailment relations helpexplainthe cancelability data in(). Since should
is logically weaker than must, using should is expected to generate a scalar implica-
ture. As is usual in scalar implicatures, Grices frst Quantity maximcan be exploited
to yield an upper-bounding implicature specifcally, from the use of should that
for all I know, or for all I am willing to presuppose for the purposes of the conversa-
tion, not must (see, e.g., Hov 1,i, 18, G.zu.v 1,, G.mU1 11). (Ofen in
scalar implicatures hearers are also licensed in taking the so-called epistemic step
(S.Uivi.u ioo), namely, from its not the case that S believes that p to S believes
that p. However, this stronger implicature will not be relevant in what follows.)
Specifcally, in using should the speaker presupposes that certain relevant applica-
bility conditions are compatible with the common ground, and implicates that, for
all she is willing to presuppose, these conditions are merely so compatible, that is,
that they do not hold throughout the worlds in the common ground. Tis scalar
implicature has the usual properties associated with implicatures. It is cancelable,
reinforceable, and tends to vanish in non-assertive contexts, as we have seen in (a),
(a), and (io), respectively. And it is suspendable, as seen in (ii).
(ii) I should help the poor. Maybe I have to.
Now lets apply the analysis in .c1U.i s1voc and sUv,Uc1ivi wi.x to our
examples in (1). Start with (1b) where it is must that is preferred. As we have
seen, afer my assertion is accepted it becomes common ground that I have no goals
18
that might outweigh my goal of traveling quickly. So, assuming that the relevant
ordering source is as given in (i1), the presupposition of must in .c1U.i s1voc
will be satisfed all of the worlds in the modal base are worlds in which the ap-
plicability condition for my goal of traveling quickly holds and your must claim
will be true. Conversely, the relative infelicity of should follows from independent
pragmatic principles. Since must is true and felicitous in this context, using the
weaker should would violate a Gricean quantity maxim. Te situation is precisely
parallel to the case of subjunctive conditionals in which it is common ground that
the antecedent holds an unsurprising point given our intuitive gloss of the truth
conditions in sUv,Uc1ivi wi.x.
Next, consider (1a). Afer my assertion is accepted it is still not accepted in
the common ground that the applicability condition for my goal of traveling quickly
holds. However, it is at least compatible with the common ground that it does. So,
since the presupposition of should but not must is satisfed, our account correctly
predicts that the weak necessity should is preferred.
Tough we have said that should is preferred in (1a), it is important to see
that in certain contexts must may be acceptable. A speaker may use must in a
context in which it is not common ground which applicability conditions hold if
she has the authority to make this common ground by her utterance. Tis is, of
course, a familiar kind of accommodation (Liwis 1,b). Suppose Alice, a young
teenager, who wants to go to Harlem and is considering with her mother, Martha,
about whether to take the A train or the C train. Te A train will be quicker, but
the C train will be safer. Martha takes the safety of her child to be of paramount
importance, though this value is not common ground between them. Nevertheless
Martha can felicitously say to Alice:
(i) You must take the C train, not the A train.
By asserting (i), Martha may, given her authority in the context, make it common
ground that the goal of traveling safely is to take priority over the goal of traveling
quickly, and thus that the applicability condition of the former but not the latter
norm holds. Tat is, she may have the authority to make Alice accommodate. Such
special contexts notwithstanding, in contexts in which the speaker does not have
this requisite authority, should will be preferred to must.
1
Bonus tracks
In sum, the meanings given in .c1U.i s1voc and sUv,Uc1ivi wi.x make in-
tuitively correct predictions concerning various semantic and pragmatic properties
of weak and strong necessity modals. Tis analysis captures certain truth condi-
tional entailment relations between should and must sentences, and makes cor-
rect predictions concerning howsubtle shifs in context can aect the relative felicity
of should and must. Like Rubinsteins account, it captures the intuitions driving
von Fintel and Iatridous account while maintaining simpler truth conditions and
better illuminating the relation between the weak/strong necessity modal distinc-
tion and the common ground. Further, it can do so without positing that weak ne-
cessity modals have a conventional signal of the (seemingly) questionable sort that
Rubinstein describes. Te putative signal that the truth of Should relies on non-
common-ground assumptions is derived from the truth conditions for should and
general pragmatic principles the same pragmatic principles governing the felic-
ity of subjunctive conditionals, the truth conditions for which provide a model for
those of should.
In closing, I would like to mention several additional benefts of our account.
First, a surprising data point highlighted in voFi1ii &I.1viuoUioo8 is that in a
wide variety of languages weak necessity modals like ought are formed by combin-
ing must with counterfactual morphology.
1o
It is speculated that the counterfac-
tual marking is co-opted here in a somewhat meta-linguistic kind of way: if we were
in a context in which the secondary ordering source was promoted, then it would
be a strong necessity that (p. 1). One might worry that, absent an explanation
of what makes it the case about a speaker that she is counterfactually promoting a
secondary ordering source, this story about howthe counterfactual marking is rele-
vant is ad hoc. Te analysis oered here may be better in this respect. Te meaning
for weak necessity modals given in sUv,Uc1ivi wi.x is, as the name suggests,
explicitly counterfactual (or, more generally, subjunctive). Roughly, Ought is
true i were such-and-such applicability conditions to hold, would be a necessity.
Ought brackets whether certain norms, goals, etc. in fact hold and is used to make
a claimabout what would be necessary if they did. Tis may yield a more natural ex-
planation for the robust cross-linguistic phenomenon of expressing weak necessity
through counterfactual marking on a strong necessity modal.
Second, though we have been focusing primarily on teleological readings of
1o
As they note, English ought fts this pattern historically: It was formerly the past subjunctive
of the verb owe (ioo8, p. 1ion.ii).
io
modals, the proposed analysis generalizes across modal favors perhaps surpris-
ingly, even to epistemic modals. Tough many authors have claimed that Should
on its epistemic reading expresses that is probable,
11
the data strongly suggests that
weak necessity modals are unlike strong necessity modals in this respect.
(i) a. Alice should be here by now, but she isnt.
b. -Alice must be here by now, but she isnt.
In credence talk, Must seems to commit the speaker to a high credence in ,
whereas Should does not seem to commit the speaker to any credence in (cf.
Sw.so io1i).
Our analysis has a natural explanation for this. It is common to view the or-
dering source used in the interpretation of epistemic modals as stereotypical, or as
consisting of propositions describing the usual or expected course of events (see,
e.g., Kv.1ziv 181a, pp. ,,, 11, pp. oo,). Just as in the case of the con-
ditional applicability of norms, goals, desires, etc., what is expected given certain
evidence can depend on whether certain conditions hold; evidential relations be-
tween propositions are also typically hypothetical, conditional on things being nor-
mal in the relevant respects. Simplifying quite a bit, in the case of (i) the ordering
source might contain a proposition E like that Alice arrives by time t

if she leaves
her house at time t, trac conditions are thus-and-so, she does not get lost, pirates
do not kidnap her along the way, and so on more generally, if no relevant ab-
normal or unexpected events occur. Tus in order for the proposition that Alice
arrives by such-and-such time to be an epistemic necessity with respect to this or-
dering source, the modal base must be restricted to worlds inwhichthis applicability
condition holds. If the modal base is not restricted in this way that is, if it is not
common ground that the applicability conditions of E hold should will be felic-
itous and it need not be epistemically necessary with respect to the actual common
ground that Alice is here by now. Te necessity of Alices being here by now only
follows from the here counterfactual assumption that conditions are normal in the
relevant respects hence the consistency of (ia). However, if the modal base is
restricted in this way that is, if it is common ground that the applicability condi-
tions of the relevant evidential norm hold must will be felicitous and it will be
epistemically necessary that Alice is here. So we derive that if Alice must be here by
now is felicitous and true, it is incompatible with the common ground that Alice is
not here hence the inconsistency of (ib).
11
See, e.g., Siom.1,o, Hov1,i, 18, Wiv1uiimiv 1,i, Wiucwoouiooo, Fii.v ioo,
L.ssi1iv io11.
i1
Actually, this last stepis tooquick. Whereas modal bases must be realistic they
can only contain propositions that are true at the world of evaluation there is
no similar constraint on ordering sources. Even if the modal base is restricted to
worlds in which the applicability conditions of E hold and so, even if must is fe-
licitous all that follows from the truth of (ib) is that the world is not among the
most normal or expected worlds if Alice is not here by now.
1i
In reply, one option is to say that Must , unlike Ought , entails that the world
of evaluation is among the
g(w)
best worlds. But this might seem to be too strong.
Even if Alice must be here by now entails that Alice is here now, it does not entail
that the world is among the most normal or expected worlds. All that follows is that
the worldis stereotypically or expectedly ideal with respect to the evidential normE. In
response, perhaps we might strengthen our truth conditions for Must by adding
an extra condition that the world of evaluation be in D(
f(w),
g

(w)
), where g

(w)
is the smallest subset of g(w) with respect to which Must is true. Tis condition
would not fgure in the truth conditions for Should .
Tough this might capture our data, there are good reasons to think that we need
not complicate our truth-conditions for must. An alternative is to follow Fv.x
1o, GiUv1s &v.uiv S.u1 1, and HUi1ixioo,a,b, among others, and say
that ordering sources may sometimes only contain propositions describing certain
contextually salient norms, regularities, goals, etc. which, indeed, may be quite
fewrather than every proposition of the kind in question. For example, suppose
the ordering source g
e
(w) only contains the proposition E. Assuming Alice must be
here by now is felicitous and true, it follows from this ordering source and the fact
that the modal base is restrictedtoworlds inwhichthe norms applicability condition
holds that the actual world is among the
g
e
(w)
best worlds. Tough the world of
evaluation need not be in general stereotypically or expectedly ideal, as far as the
interpretation of must goes in (ib) with respect to the norm E, it is; distinctions
are only made among worlds in the modal base depending on whether Alice is here
by now in them. Tus, if must is felicitous and You must go to confession is true,
it is incompatible with the common ground that Alice is not here by nowhence
the inconsistency of (ib).
1
Data like (i,) suggest that, surprisingly, even for deontic readings of the modals,
Must , unlike Should, commits the speaker to a highcredence in. Must
1i
Tanks to Dan Greco for discussion.
1
For alternative pragmatic explanations that appeal to the dierent performative eects of weak
and strong necessity modals, see Ni. ioo,; Sw.so ioo8, pp. 1io1io; Pov1iv ioo,
pp. 1o, 181o.
ii
appears to be inconsistent on deontic readings as well.
1
(i,) a. You should go to confession, but you wont (/but I know you wont).
b. -You must go to confession, but you wont (/but I know you wont).
Of course, obligations can go unfulflled. However, speakers appear to rule out this
possibility at least for the purposes of the conversation when expressing obli-
gations with must.
Our explanations of the data in (i) carry over quite elegantly to the deontic
case.
1,
Whether the norm(s) in question is (are) evidential or deontic, speakers, at
least for the purposes of the conversation, rule out the possibility that it (they) will
go unsatisfed in utterances of Must . In terms of our second strategy above, in the
case of (i,b) the ordering source g
d
(w) might only contain the proposition D that
you go to confession i C, for some relevant applicability conditions C. Assuming
You must go to confession is felicitous and true, it follows fromthis ordering source
and the fact that the modal base is restricted to C-worlds that the world of evalu-
ation is among the
g
d
(w)
best worlds. Tough the world of evaluation need not
be in general deontically ideal, as far as the interpretation of must goes in (i,b)
with respect to the norm D, it is; distinctions are only made among worlds in the
modal base depending on whether you go to confession in them. Tus, as in the
epistemic case above, if You must go to confession is felicitous and true, it is in-
compatible with the common ground that you do not go to confession hence the
inconsistency of (i,b). But if the modal base is not restricted to C-worlds, or if the
ordering source contains further propositions that are not satisfed in the world of
evaluation, should will be felicitous and it need not be deontically necessary with
respect to the actual common ground that you go to confession. Te necessity of
your going to confession only follows from the here counterfactual assumption that
C holds hence the consistency of (i,a).
In this way, the analysis in this paper can provide a uniform explanation for the
behavior of weak and strong necessity modals that generalizes across modal favors.
1
See esp. Wiviv ioo, pp. 1i1,; Ni. ioo,; Pov1iv ioo, pp. 1o, 181o; cf., e.g.,
Nowiii-Smi1u1,, pp. io,io8, Limmo1oi, pp. 181,o, Wiv1uiimiv1,i, H.vm.1,,
Lvos 1,,, Wiiii.ms 181, P.imiv ioo1.
1,
Tanks to Bob Stalnaker for discussion.
i
References
Bi.viv, D.viu. ioo1. Presupposition and assertion in dynamic semantics. Stanford:
CSLI Publications.
Bi.viv, D.viu, Cv.ici Roviv1s, M.uv Simos, & JUui1u Tou.Usiv. iooa.
Not just projecting. MS: UT Austin, OSU, CMU, and OSU.
Bi.viv, D.viu, Cv.ici Roviv1s, M.uv Simos, & JUui1u Tou.Usiv. ioob.
Presupposition, conventional implicature, and beyond: Aunifed account of pro-
jection. In Proceedings of new directions in the theory of presupposition. Toulouse:
ESSLLI.
Cuiivcui., Gi.vo & S.iiv McCoiii-Gii1. 1o. Meaning and grammar.
An introduction to semantics. MIT Press: Cambridge.
Coviiv, Bviuci1. iooo. What should should mean: MS, CNRS/Universit Paris 8.
Fii.v, S1ivui. ioo. What ought probably means, and why you cant detach it.
Synthese r,,, o,8.
vo Fi1ii, K.i. ioo. Would you believe it: Te king of France is back! Presup-
positions and truth-value intuitions. In M.vc. Riimiv &Ai BizUiuiuoU1
(Eds.), Descriptions and beyond, pp. 1,1. NewYork: Oxford University Press.
voFi1ii, K.i &S.vii I.1viuoU. ioo,. What to do if you want to go to Harlem:
Anankastic conditionals and related matters. MS, MIT.
vo Fi1ii, K.i & S.vii I.1viuoU. ioo8. How to say ought in foreign: the
composition of weak necessity modals. In J.cqUiiii GUivo & J.cqUiiii
Lic.vmi (Eds.), Time and modality, pp. 11,11. Springer.
v. Fv..ssi, B.s C. 1,. Values and the hearts command. Te journal of Phi-
losophy ,o, ,1.
Fv.x, Ai11i. 1o. Context dependence in modal constructions. Ph.D. thesis,
University of Stuttgart.
G.mU1, L.T.F. 11. Logic, language, and meaning, Volume r. Introduction to logic.
Chicago: University of Chicago Press.
i
G.zu.v, Giv.iu. 1,. Pragmatics. Implicatures, presupposition, and logical form.
New York: Academic Press.
GiUv1s, B.v1 & Rov v. uiv S.u1. 1. Domain restriction. In Pi1iv Boscu
& Rov v. uiv S.u1 (Eds.), Focus. Linguistic, cognitive, and computational per-
spectives, pp. io8ii. Cambridge: Cambridge University Press.
H.vi, R.M. 1,i. Te language of morals. Oxford: Oxford University Press.
H.vm., Giiviv1. 1,. Review of Roger Wertheimer, Te signifcance of sense.
Meaning, modality, and morality. Te Philosophical Review 8:, i,i.
H.vm., Giiviv1. 1,,. Reasons. Crtica. Revista Hispanoamericana de Filosofa
,, 1,.
H.vm., Giiviv1. 1. Stringency of rights and ought. Philosophy and Phe-
nomenological Research ,, 18118,.
Hov, L.Uvici R. 1,i. On the semantic properties of logical operators in English.
Ph.d. thesis, University of California, Los Angeles.
Hov, L.Uvici R. 18. A natural history of negation. Chicago: University of
Chicago Press.
HUi1ix, J.ixi. ioo,a. Analyzing anankastic conditionals and suciency
modals. In S. Bi.uo, L. Vici1i, & E. Scuooviimmiv (Eds.), Proceedings of
Console r, pp. 1,1,o. University of Leiden.
HUi1ix, J.ixi. ioo,b. Anankastic conditionals and salient goals. In Em.v
M.iiv, Covii B.vv, & J.ixi HUi1ix (Eds.), Proceedings of Sinn und Be-
deutung , pp. 1o1,.
Jois, Auviw J.I. & Icm.v Pov. 18o. Ought and must. Synthese oo, 8.
K.1, Imm.Uii. 1,o. Groundwork of the metaphysics of morals. P.1o, H.J.
(Trans.), New York: Harper & Row, rd edn.
K.v11Ui, L.Uvi. 1,. Presupposition and linguistic context. Teoretical lin-
guistics r, 1811.
Kv.1ziv, Aciiix.. 1,,. What must and can must and can mean. Linguistics
and Philosophy r, ,,,.
i,
Kv.1ziv, Aciiix.. 1,. Conditional necessity and possibility. In R.iiv
B.Uivii, Uvs Ecii, & Avim vo S1icuow (Eds.), Semantics from dierent
points of view, pp. 11,1,. New York: Springer.
Kv.1ziv, Aciiix.. 181a. Te notional category of modality. In H.s JUvci
Eixmiviv & H.is Riisiv (Eds.), Vords, worlds, and contexts. ^ew approaches
in word semantics, pp. 8,. Berlin: de Gruyter.
Kv.1ziv, Aciiix.. 181b. Partition and revision: Te semantics of counterfac-
tuals. journal of Philosophical Logic ro, io1i1o.
Kv.1ziv, Aciiix.. 18. An investigation of the lumps of thought. Linguistics
and Philosophy r:, oo,o,.
Kv.1ziv, Aciiix.. 11. Modality/Conditionals. In A. vo S1icuow &
D. WUuiviicu (Eds.), Semantics. An international handbook of contemporary
research, pp. oo,o. Berlin: de Gruyter.
Kv.1ziv, Aciiix.. iooi. Facts: Particulars or information units: Linguistics and
Philosophy :,, o,,o,o.
L.ssi1iv, D.iii. io11. Measurement and modality. Te scalar basis of modal se-
mantics. Ph.D. thesis, New York University.
Limmo, E.J. 1oi. Moral dilemmas. Philosophical Review ,r, 11,8.
Liwis, D.viu. 1,. Counterfactuals. Cambridge: Harvard University Press.
Liwis, D.viu. 1,a. A problem about permission. In E. S..vii, R. Hiivii,
I. NiiiiUo1o, & M.P. Hi1ixx. (Eds.), Essays in honour of jaakko Hintikka, pp.
1o1,,. Dordrecht: Reidel.
Liwis, D.viu. 1,b. Scorekeeping in a language game. journal of Philosophical
Logic 8, ,.
Liwis, D.viu. 181. Ordering semantics and premise semantics for counterfactu-
als. journal of Philosophical Logic ro, i1,i.
Lvos, Jou. 1,,. Semantics, vol. i. Cambridge: Cambridge University Press.
Ni., Diiiv. ioo,. Two puzzles about deontic necessity. In JoG.,iwsxi, V.ii-
1ii H.cqU.vu, Biv.vu Nicxii, & Si1u Y.ici (Eds.), ^ew work on modal-
ity. MIT working papers in linguistics, vol. ,1, pp. 11,8.
io
Nowiii-Smi1u, P.H. 1,. Ethics. Baltimore: Penguin Books.
P.imiv, F.R. ioo1. Mood and modality. Cambridge: Cambridge University Press,
ind edn.
Pov1iv, P.Ui. ioo,. Imperatives and modals. ^atural Language Semantics r,,
,18.
Pov1iv, P.Ui. ioo. Modality. Oxford: Oxford University Press.
Ross, W.D. 1o. Te right and the good. Oxford: Oxford University Press.
RUvis1ii, Av.1. io11. Projective signals of weak necessity modals. In Gvicovv
Kiivs1i.u (Ed.), Proceedings of ESSLLI :orr. Vorkshop on projective content, pp.
111,,. Ljubljana.
RUvis1ii, Av.1. io1ia. Figuring out what we ought to do. Poster for the Penn
Linguistics Colloquium (PLC o).
RUvis1ii, Av.1. io1ib. Roots of modality. Ph.d. thesis, University of Mas-
sachusetts.
S.Uivi.u, Uii. ioo. Scalar implicatures in complex sentences. Linguistics and
Philosophy :,, o,1.
Su.o, Biv. 1,o. On the two kinds of presupposition in natural language.
Foundations of language r,, i,i.
Siix, Aiix. io1i. Deontic conditionals. MS, University of Michigan.
Siom., A.vo. 1,o. ought and better. Mind ,, 8,.
S1.i.xiv, Roviv1. 1,. Pragmatic presuppositions. In S1.i.xiv (1), pp.
,oi.
S1.i.xiv, Roviv1. 1,8. Assertion. In S1.i.xiv (1), pp. ,8,.
S1.i.xiv, Roviv1. 1. Context and content. Essays on intentionality in speech
and thought. Oxford: Oxford University Press.
S1.i.xiv, Roviv1. iooi. Common ground. Linguistics and Philosophy :,,
,o1,i1.
i,
Sw.so, Evic. ioo8. Modality in language. Philosophy Compass , 111io,.
Sw.so, Evic. io11. On the treatment of incomparability in ordering semantics
and premise semantics. journal of Philosophical Logic ,o, o,1.
Sw.so, Evic. io1i. Te application of constraint semantiics to the language of
subjective uncertainty. journal of Philosophical Logic pp. 1i8.
Vii1m., Fv.x. 1,o. Prejudices, presuppositions, and the theory of condition-
als. In Jivoi Gvoiiui,x & M.v1i S1oxuoi (Eds.), Amsterdam papers in
formal grammar, vol. 1, pp. i8i81. Central Interfaculteit, University of Ams-
terdam.
Wiucwoou, R.ivu. iooo. Te meaning of ought. In RUss Su.iiv-L.u.U (Ed.),
Oxford studies in metaethics, vol. r, pp. 1i,1oo. Oxford: Oxford University Press.
Wiviv, Tuom.s A. ioo. Deducing the future and distinguishing the past. Temporal
interpretation in modal sentences in English. Ph.d. thesis, Rutgers University.
Wiv1uiimiv, Rociv. 1,i. Te signifcance of sense. Meaning, modality, and moral-
ity. Ithaca: Cornell University Press.
Wiiii.ms, Biv.vu. 181. Practical necessity. In Moral luck, pp. 1i11. Cam-
bridge: Cambridge University Press.
i8