The linguistic basis for propositions

Peter van Elswyk, University of Wisconsin–Milwaukee

Penultimate draft ⋅ Forthcoming in Routledge Handbook of Propositions

Propositions are traditionally regarded as performing vital roles in theories

of natural language, logic, and cognition. This chapter offers an opinion-
ated survey of recent literature to assess whether they are still needed to
perform three linguistic roles: be the meaning of a declarative sentence in
a context, be what is designated by certain linguistic expressions, and be
the content of illocutionary acts. After considering many of the relevant
theoretical choice-points, I suggest that there remains a linguistic basis for
propositions, but not for some of the traditional reasons.

1 Introduction

Proposition is a theoretical term. It names an entity whose existence is inferred as

opposed to a kind that is directly observed.1 Theories of natural language, logic,
and cognition require an entity to fulfill (R1) - (R8).

(R1) Be the meaning of a declarative sentence linguistic roles

(R2) Be the designation of a linguistic expression
(R3) Be the content of illocutionary acts
(R4) Be the relata of entailment relations logical roles
(R5) Be the bearer of alethic properties
(R6) Be the bearer of modal properties
(R7) Be the objective of cognitive attitudes cognitive roles
(R8) Be the content of perception

Not everybody agrees that all of the roles above are required by a successful
theory to have some entity or other fill them. Perhaps (R7) is dispensable
because our best theory of cognition does not require an object for attitudes
1
A question to raise is whether we should be realists about propositions as unobserved entities.
See Simchen (forthcoming) for a defense of instrumentalism about structured propositions,
Armour-Garb and Woodbridge (2012) for a pretense-based fictionalism, and Ball and Rabern
(2018, §1.2.2) for a survey of anti-realist attitudes towards meanings generally. In §3, I identify
some initial problems with non-realist views.
like belief. Nevertheless, proposition names the kind of entity which performs or
is suited to perform roles like those enumerated.
The roles above are the primary roles. On that assumption that one entity
performs all or most of them, secondary roles are derivable from what must be
true of whatever entity performs the primary roles. For example, propositions
must be abstract objects that are not spatially located. They might be temporally
located depending on what it takes for them to come into existence and persist.
Let’s call this the functional characterization.2 Beginning with theoretical
roles is a standard way propositions are motivated. Many justify their use of
the label proposition according to whether their subject matter fulfills the roles.
Horwich (1998, 82), for example, writes that “it is a matter of stipulation to
call the entities to which [that-clauses] refer, ‘propositions’.” After or before
considering various roles, Schiffer (2003, 14) summarizes, “From all this we
may conclude. . . that things believed are what philosophers nowadays call
propositions” and Merricks (2015, 21) writes “I conclude that we should apply
the label ‘propositions’.”
By itself, the functional characterization clarifies little about what propo-
sitions are. They may be sui generis representational entities, or something
more ontologically familiar like a fact, property, set, or abtract type. What
propositions explain within our theories is not clarified either. Maybe we
require propositions for (R2) but propositions are otherwise explanatorily idle.
However, what is highlighted by characterizing propositions functionally is how
to evaluate whether there are any propositions: by considering their putative roles
in leading theories.
In this chapter, I provide an opinionated survey on the linguistic roles.
For each of (R1), (R2), and (R3), I consider two questions: whether the
role is required by a successful linguistic theory, and whether a proposition
is still needed to perform the role. Having a linguistic basis for propositions
requires an affirmative answer to both questions. I explore whether, as matters
presently stand, we have continued cause to posit the existence of propositions
in explanations of natural language. Though I will suggest that there remains
a linguistic basis for propositions, the reader will see it is not for many of the
traditional reasons.
Along the way, various non-propositionalisms about the roles will be en-
countered. I distinguish easy from hard non-propositionalism.3 Easy non-
propositionalists are spooked by propositions for metaphysical and/or episte-
mological reasons. They regard the spookiness of propositions as enough to do
2
See Crawford (2014) for a nearby distinction between conceptions of propositions that are
metaphysically lightweight as opposed to metaphysically heavyweight.
3
The easy/hard distinction is borrowed from Colyvan (2010) who draws a similar distinction
about the metaphysics of mathematical objects.

2
without them. As a result, they do not bother to show how explanations involving
the theoretical roles can be developed without positing propositions. In contrast,
hard non-propositionalists do attempt to show how such explanations can be
developed. I often ignore easy non-propositionalisms in what follows. Absent
an explanation for why the roles are no longer necessary and/or an illustration
for how other entities perform the roles, the lingusitic basis for propositions
remains.
A fews caveat are owed before the survey begins. First, I occasionally opine
more than argue. Theoretical choice-points relevant to whether a role provides
a basis for propositions will be identified, a choice will be suggested, and yet I
will not offer much motivation for my suggestion. Second, I survey quickly.
Since I canvass many views, exposition will not run as deep as it would if I were
just focusing on one view. Finally, I make omissions. For example, I will not
say much about dynamic semantics.4 Each limitation is a consequence of the
chapter’s modesty. I do not intend to settle whether propositions are required
for the linguistic roles. I aim to highlight the choice-points at stake for the benefit
of future work on the basis for propositions, and identify a plausible path through
the choice-points.

2 The meaning of a declarative

2.1 Assigning entities

A semantic theory explains meaning facts. Some meaning facts are particular.
You cannot felicitously use that to refer to an object you are holding in your
hands—you have to use this instead. Other meaning facts are general. The
following principle illustrates generality.

principle of compositionality
The meaning of a complex expression e in a language L is deter-
mined by the meanings of e’s parts and the way they are combined
according to L’s syntax.

The success of semantic theories can be compared by how well they explain par-
ticular or general meaning facts. A semantic theory, for example, which explains
4
Such an omission might irritate some readers because dynamic theories provide rich resources
for rethinking (R1), (R2), and (R3) along with the logical roles. Though what I will say below
has bearing on whether propositions persist as entities for which a dynamic semantics requires
existence, an adequate discussion is not conducive to the survey nature of this chapter. To my
knowledge, no such discussion has been offered that covers all of the pertinent theoretical roles.
For a discussion of truth in dynamic semantics, see Stokke (2014). Stojnic (forthcoming) attends
to how to understand propositional content in a dynamic semantics. Murray and Starr (2018)
consider to understand illocutionary force alongside dynamic semantics.

3
the meanings of complex expressions in a manner that violates compositionality
is a non-starter. It fails to explain an important fact.
We can divide semantic theories into entity-assigning and entity-free
theories. Both explain how expressions have meaning. They differ in whether to
have meaning is to have a meaning. Accoringly, entity-assigning theories reify
meanings. For an expression e to have meaning is for e to have a particular
entity assigned as its meaning by the semantic theory. In contrast, entity-
free theories are ontologically austere. For an expression e to have meaning
is for e to be characterizable in a certain way by the semantic theory where
such characterization does not require the existence of an entity to be assigned.
Whether (R1) is necessary as a role depends on whether our leading semantic
theories are entity-assigning or not.
A historically important entity-free approach is the truth-theoretic se-
mantics inaugurated by Davidson (1967). On this approach, the meaning of
a sentence S is characterized by the conditions under which S is true in the
semantic theory. Such characterizing is done with the biconditional (T) where
S is replaced with a sentence in the object language, and p is replaced with a
sentence in the metalanguage of the semantic theory. Following Davidson, we
can call what does the characterizing an axiomatic truth theory.

(T) S is true if and only if p.

Though an axiomatic truth theory does not explain meaning, Davidson’s pro-
posal was that putting constraints on one does. Truth-theoretic semantics
therefore pose a challenge to an (R1)-based reason for positing propositions.
Were we to adopt a semantic theory like Davidson’s, there would be no need for
propositions because we have no need for any entity to perform (R1).
Truth-theoretic semantics faces significant explanatory limitations. Many
meaning facts are not explainable with truth or related notions. There are
clause types other than declarative that do not have a meaning assessable for
truth or falsity (e.g. interrogatives, imperatives, exclamatives), and words or
morphemes with meaning that do not contribute a declarative’s truth-conditions
(e.g. slurs, honorifics, discourse markers). No similar limitation is faced by a
semantics which freely assigns entities. The difference between clause types,
for example, can be captured by assigning different entities to them. The
dedicated defenders of truth-theoretic semantics are aware of these limitations
and have made limited attempts to overcome them.5 But it would be a stretch
to classify truth-theoretic semantics as being equal to or greater than entity-
assigning theories in explanatory success.
5
See Lepore and Ludwig (2007, Ch.12), for example. But even that dynamic Davidsonian duo
does not agree on how to capture non-declaratives. See Starr (2014) for some recent discussion
of the problem with a truth-theoretic approach to mood.

4
 Another entity-free theory to mention is internalist semantics. It departs
from the externalism familiar from Putnam (1975) and Burge (1979) on
which meanings have extensions that are determined externally. Though such
departures take many shapes, internalist semantics, as I have in mind, originates
with the skepticism towards reified meanings found in the work of Noam
Chomsky.6 A particularly forceful defense of it is given by Pietroski (2018).
Meanings are, for him, instructions for building or accessing concepts. The
instructions remain fairly simple in the process of semantic composition by
being massively conjunctive. Importantly, concepts are not entities that are
assigned as meanings to words or phrases.
As with truth-theoretic semantics, internalist semantics ensures that propo-
sitions are not needed to perform (R1) because nothing is required for the job.
However, it importantly differs from truth-theoretic semantics in its explanatory
promise. Where truth-theoretic semantics is limited to explaining meaning
with notions based or oriented around truth, internalist semantics faces no such
limitation. Clause types other than declaratives, for example, can be given
instructional meanings as Lohndal and Pietroski (2011) show for interrogatives.
In this chapter, I do not have much more to say about the philosophical
viability of internalist semantics partly because it would take us too far afield
into the philosophy of mind.7 But there is a sociological point to make. Few
philosophers and linguists explain meaning facts with an entity-free semantics.
Most advances made in the last 50 years in understanding the meaning facts of
natural languages were overwhelmingly made while assigning entities of some
kind. It might be, of course, that these advances could have been made with
an entity-free semantics, or that such advances can be mimicked with one. As
a matters stand, entity-assigning theories have considerably more explanatory
coverage of meaning facts.

2.2 Assigning propositions

But what entity performs (R1)? Contemporary semantic theories are over-
whelmingly model-theoretic.8 The meaning of an expression is given by an
interpretation function J⋅K that assigns an entity from a model to an expression of
natural language. Those entities are drawn from a domain that consists of basic
and non-basic entities. Basic entities are individuals like you, me, and truth-
values. Non-basic entities are typically complex functions to and from either
6
Such skepticism can be found in many of Chomsky’s writings, but see especially Chomsky
(1995) and Chomsky (1996).
7
For a recent defense of externalism, though, see Yli-Vakkuri and Hawthorne (2018).
8
Accordingly, advocates of truth-theoretic semantics often take issue with model-theoretic
semantics. See Lepore (1983) and Glanzberg (2014) for discussion.

5
basic entities or other non-basic entities. Within model-theoretic semantics, the
meaning of a declarative is therefore a proposition if a proposition is assigned by
J⋅K.
Within a standard intensional semantics, the meaning of a sentence is an
intension: a function from worlds or world/time pairs to truth-values. Many
therefore follow Montague (1974) and use proposition to name those sets of
worlds which value sentences to true.9 Others ocassionally use proposition to
name whatever entity is assigned by J⋅K to a declarative. These different uses
can cause confusion. In adopting a functional characterization, I have reserved
proposition for what is apt to fulfill the primary roles (R1) through (R8). But, for
the purpose of this section, I will assume that a proposition is a set of worlds or
a similar entity.
Following Rothschild (2013, 49-50), various non-propositionalisms can be
distinguished. There is blanket non-propositionlism that assigns an entity other
than a proposition to every declarative. Then there are non-propositionalisms
that are more selective. domain-driven non-propositionalism maintains that
declaratives in a specific domain of discourse do not have propositions for
meanings. Declaratives about what is morally right or wrong, for example, might
express speaker attitudes as opposed to have propositions for meanings. Finally,
linguistic-driven non-propositionlism holds that declaratives containing certain
expressions cannot be assigned propositions because of facts local to the expres-
sions.
At this particular point in the history of semantics, non-propositionalisms
of both blanket and selective variety abound. One merely has to open the latest
issue of a semantics journal like Journal of Semantics or Natural Language Semantics
to find a proposal where J⋅K doles out entities that are arguably unfit to perform
the other theoretical roles. From such pluralism about entity assignment, it is
reasonable to draw the conclusion that successful semantic theories do not need
to have propositions perform (R1). An array of entities can be assigned, and it
will not get in the way of explaining various meaning facts. Let’s call this the
pluralism problem for basing the existence of propositions on (R1).
Note that blanket propositionalism does not need to be true for (R1) to
provide a reason to posit propositions. As long as some declaratives still have
propositions assigned by J⋅K, a reason remains. At best, domain-driven and
linguistic-driven non-propositionalisms only weaken the need for propositions.
Blanket non-propositionalisms pose the most serious challenge. But at least two
variants of (R1) might be true even if blanket propositionalism is not.
9
An intensional semantics can be set-up in a different way that allows propositions to be
something else. For example, see Thomason (1980), Muskens (2005), and Pickel (forthcoming).
Pollard (2015) offers a hyperintensional semantics which is able to remain agnostic on what
propositions are.

6
 (R1) Be the meaning of a declarative sentence
(R1*) Be a component of the final meaning of a declarative
(R1**) Be an essential part of the meaning of a declarative

(R1*) makes room for the meaning of a declarative to be a complex entity that
includes a proposition as one of its components. The next variant is (R1**).
It makes room for the meaning of a declarative to be something other than a
proposition but something which mandatorily composes with a proposition.
Let’s see some examples of each variant role. There are many semantic
theories that do not require propositions for (R1) but do for (R1*). Using p as a
variable for whatever a proposition is by the light of the semantic theory, there
are theories where the meaning of a declarative is {p}, a singleton proposition,
and theories where what is assigned by J⋅K is ⟨p, . . . ⟩, an ordered pair with a
proposition as a coordinate.10 Strictly speaking, theories such as these do not
require propositions to be the meaning of a declarative sentence. But it is as
not as if propositions have disappeared. They merely surfaced elsewhere in the
semantic theory.
One example of the many theories that retains (R1**) is the commitment
state semantics of Krifka (2015). In such a theory, the meaning of a declarative
is a proposal to change what the speaker’s commitments are. What is assigned
as a meaning to a declarative is then a complex function from one state of
commitments to another. But that entity essentially involves a proposition
because what a speaker is committed to are propositions. So in the semantic
composition an operator takes a proposition to convert it into a function that
behaves as a proposal to add that proposition to the speaker’s commitments.
Propositions are unnecessary yet again for (R1), but remain indispensable to
perform (R1**).11
In some cases, selective non-propositionalism may also retain a variant role.
Simplifying for our purpose, declaratives syntactically decompose into a tense
phrase that can be dominated by an array of additional phrases. Suppose a
proposition is the meaning of a tense phrase. Then the meaning of a declarative
will be a proposition if and only if no phrase dominating the tense phrase hosts
an element that converts the proposition contributed by the tense phrase into
10
For example, see multidimensional semantic theories in the style of Kartunnen and Peters
(1979), Potts (2004), and McCready (2010).
11
Retaining (R1**) is also common to theories that move to blanket non-propositionalism from
linguistic-driven non-propositionalism. For example, Moss (2018) regards declaratives with
epistemic vocabulary as requiring probabilistic content even though other declaratives do not.
To have a unified semantic theory, she introduces a type-shifter that converts a proposition had
by the meaning of an epistemic-free declarative to a set of probability spaces. Strictly speaking,
the meaning of every declarative is therefore a set of probability spaces but (R1**) lingers for
the sentences that lack epistemic vocabulary.

7
a different entity. In this light, consider non-propositionalism about epistemic
modals. The meaning facts surrounding words like might and must have led many
to adopt theories where the meaning assigned to a declarative like Durian might
be pungent is not a proposition. But epistemic modals are hosted by a phrase that
usually dominates the tense phrase (Cinque, 1999). For many semantic theories
of epistemic modals, modals convert the proposition contributed by the tense
phrase into a different entity. Such theories falsify (R1) for declaratives with
modals, but retain (R1**) given what is assigned.
Here’s the upshot. In considering semantic theories that are non-propositional
across the board, it is important to assess whether (R1*) and/or (R1**) linger
as roles. If they do, the need for propositions has minimally wavered. A
proposition may not be the meaning that is assigned to a declarative by a leading
semantic theory, but, whatever entity is assigned, propositions are still essential
to that entity’s identity. By itself, pluralism therefore carries no consequence
for whether an explanation of a declarative’s meaning requires the existence of
propositions. What would dispense with propositions would be a blanket non-
propositionalism that leaves no roles behind and which has explanatory coverage
unrivaled by even a partially propositional theory. I submit that no such theory
presently exists.
A thorough examination of the pluralism problem would require us to
compare propositional and non-propositional theories side-by-side with respect
to how they explain general and particular meaning facts. I do not offer that here.
Side-by-side comparisons can be found elsewhere for meaning facts related to
a few particular expressions.12

3 The designation of a linguistic expression

3.1 Designating propositions

I turn now to (R2). In my discussion I have to take some side or other about
how to explain meaning facts. Accordingly, I assume a textbook entity-assigning
semantics in the tradition of Montague (1974).
Let’s call the underlined terms entity terms and the sentences in which they
occur entity sentences.

(1) What was said by Andrew is true / believable. free relative

(2) Andrew believes that durian is pungent. that-clause
12
Consult King (2007) and Ninan (2012) for a discussion of how propositional and non-
propositional semantic theories handle tense. Among numerous others, see Yalcin (2007), Moss
(2018), Stojnic (forthcoming), and Mandelkern (2019) for representative perspectives on how
different theories explain the facts surrounding epistemic modality.

8
 (3) Brad thinks so too. anaphor
(4) The proposition that durian is pungent is true / believable. def-
inite
(5) Logicism is true / believable. name

The existence of propositions can be argued for with a few assumptions about
both entity terms and sentences. The traditional case relies upon the loaded
notion of a singular term. Singular terms reference a single individual.
They contrast with general terms that pick out many individuals and non-
terms like predicates. True sentences containing singular terms require the
existence of what the terms reference. Accordingly, if there are true entity
sentences featuring singular terms for propositions, propositions earn existence.
Variants of this reasoning are available. True sentences containing an existential
quantifier require the existence of a witness. Entity terms could be quantifiers
that only take propositions as witnesses. Similarly, one might maintain that the
entity terms contribute a universal quantifier with a non-empty domain that
consists exclusively of propositions.
I follow King (2002, 342) in using designation to name how entity terms are
putatively related to propositions. As he defines it,

An occurrence of expression e in sentence S designates o iff this

occurrence of e is via some semantic mechanism associated with
o and as a result S, in virtue of containing this occurrence of e,
expresses a proposition P whose truth or falsity at a circumstance
depends on the properties of o and the relations it stands in at that
circumstance.

Talk of designation allows us to bypass how precisely an entity term contributes

to the truth-conditions of a sentence such that those truth-conditions depend
on an entity being a certain way. Accordingly, it enables us to avoid discussion
of singular termhood. I take that to be a benefit. The entity terms in (1)
through (5) cross-cut a variety of syntactic categories. For example, the names
and descriptions are determiner phrases, but that-clauses are complementizer
phrases and free relatives behave like determiners but are not. Contemporary
diagnostics for identifying singular terms such as those developed by Hale (1994,
1996) only work expressions of a particular syntactic category. They do not work
for us.
What I call the entity argument assumes (A1)-(A3). (A1) and (A2) gets us
(R3), and (A3) solidifies that propositions fulfill it.

(A1) There are true entity sentences.

9
 (A2) Entity terms designate.
(A3) Propositions are what entity terms designate.

I will not discuss whether (A1) is true. It is difficult to take serious the claim
that no entity sentences are ever true. My focus in this section is whether (A2)
and (A3) are true. Both are claims about the meanings of the entity terms. The
truth or falsity of either is motivated by considering meaning facts and what a
compositional semantics of the terms needs to be to explain such facts.
Earlier, I said I would only take hard non-propositionalisms seriously.
Somewhat alarmingly, non-propositionalist discussions of (R2) frequently take
the easy road. Let me offer examples. Armour-Garb and Woodbridge (2012)
offer a pretense-based view of (R2). Propositions do not exist, but pretending
they do enables speakers to make “claims about certain complex use-features
of mental and linguistic items, and for expressing certain generalized claims
about these features that we could otherwise not express (647).” They advertise
that their view is a semantic view in that it is neither pragmatic nor merely
paraphrase. It connects sentences to truth-conditions without requiring users
of the language to be aware of the pretense.
So what are complex use-features? Their answer is that they are long-arm
conceptual roles such that “specifications and attributions of these use-features
would inevitably be extremely long, complicated and technical (650).” When it
comes to particular entity terms, schematic answers only are given. For example,
a that-clause stands in for whatever is the long-arm conceptual role that the
speaker associates to the clause. But this falls short of taking the hard road.
We are never told what such use-features are even for a simple example. Nor
are we ever given an entity-assigning semantics on which these use-features are
provided as the designation of entity terms. So their view is a mere promissory
note.
I offer an additional example. Jubien (2001) aims to do away with proposi-
tions as the intermediary between a subject and what a proposition represents.
So what plays the theoretical roles are the individuals, properties, and relations
that a proposition would represent if there were any propositions. Such a
collection of things is held together—glued, to use Jubien’s metaphor—by the
intentionality of an agent. The objects of thought for Durian is pungent is
something like the Platonic property of pungentness and the natural kind
designated by Durian. When it comes to that-clauses, Jubien claims that
they involve plural quantification over the relevant individuals, properties, and
relations. But he demurs that he “can’t go into details of plural quantification
here (57)” and never returns to give them. Without the details, though, we do
not have an explanation of the meaning facts surrounding that-clauses. Perhaps
the easy non-propositionalisms canvassed can be developed to be explanatory

10
hard non-propositionalisms. Count me skeptical.

3.2 The case for (A2)

3.2.1 Free relatives

Relative clauses fronted with a wh-expression like who or what are free if they
do not attach to a noun. Instead, they constitute an independent constituent.
Free relatives involving what are frequently used to to designate proposition-like
entities introduced in a conversation.
⎧
⎪ said ⎫ ⎪ ⎧
⎪ true ⎫⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎪
(6) What was ⎨ uttered ⎬ is ⎨ thought ⎬.
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪believed⎪
⎪ ⎪
⎪
⎪
⎩believed⎪⎭ ⎪ ⎩ ⎪
⎭
Though there is disagreement over their underlying syntax and how their
final meanings are compositionally determined, consensus has formed that free
relatives have a quantificational semantics. A free relative like what was said is
thought to have the same semantics as either a definite like the thing(s) that was
said or an indefinite such as a thing that was said. When it composes with -ever to
form whatever was said, it is widely regarded as having the semantics as a universal
quantifier like everything that was said.13 Free relatives satisfy (A2).

3.2.2 That-clauses

A traditional view is that the that-clauses in sentences like (7) and (8) are name-
like terms for propositions.

(7) Andrew believes that durian is pungent.

(8) That durian is pugnent is true.

There are two standard reasons for regarding that-clauses as designative ex-
pressions. The first is what I dub the evidence from relational attitudes.14
On a traditional analysis of attitudes, verbs like hope or believe are two-place
relations between a subject and a proposition. Since a that-clause is the apparent
complement to an attitude verb, it must designate a proposition. Another reason
is based on what I will call the evidence from valid inferences.15 Many maintain
that regarding that-clauses as referential allows us to explain inferences like (9)
and (10).
13
See Berman (1991) for an indefinite analysis, Jacobson (1995) for a definite analysis, and
Dayal (1997) for a universal analysis of -ever free relatives. A helpful introduction to their
14
semantics is provided by Šimík (forthcoming). See Schiffer (1972) and Stalnaker
15
(1987). Consult Schiffer (1972) and Bealer (1998).

11
 (9) Andrew believes that durian is pungent.
Brad believes that durian is pungent.

There is something that Andrew and Brad both believe.

(10) Andrew believes everything said by Brad about durian.
Brad said that durian is pungent.

Andrew believes that durian is pungent.

In recent years, however, both of these bits of evidence for that-clauses being
designative have been called into question.16
An increasingly discussed problem for that-clauses being designative involves
substitution. It is widely thought that two co-referring terms can be substituted
for one another without altering grammaticality or truth in an extensional
context.17 But that-clauses arguably cannot. If we assume that descriptions like
the proposition that durian is pungent designate propositions, then we can produce
substitution failures with that-clauses and descriptions.

(11) (a) Andrew fears that durian is pungent.

(b) Andrew fears the proposition that durian is pungent.
(12) (a) Andrew hopes that durian isn’t pungent.
(b) # Andrew hopes the proposition that durian isn’t pungent.

In (11b), it appears that the meaning changes after substitution. Unless Andrew
is a committed non-propositionalist, (11b) can be false while (11a) is true. A loss
of grammaticality is shown in (12b) after substitution.
Three primary responses are offered. First, take the failures as evidence
against that-clauses designating propositions.18 Second, explain the failures as
a grammatical quirk owed to the environment in which substitution fails.19
For example, King (2002) notes that the grammatical object of hope must be a
complementizer phrase. Since proposition descriptions are determiner phrases,
the loss of grammaticality in (12b) can be explained on independent grounds.
Third, deny altogether that proposition descriptions have the same meanings
as that-clauses.20 That way, substitution failure can be blamed on a meaning
difference between the terms.
16
See Bach (1997), Hofweber (2007), Rosefeldt (2008), and van Elswyk (forthcoming) for very
different diagnoses of how the evidence is not probative goes awry.
17
For discussion of intersubstitutibility, see Wright (1998), Dolby (2009), Trueman (2012,
2018), and Nebel (2019). 18 See especially Moltmann (2003). 19 See King (2002) and Nebel
(2019) for different versions of this response. 20 Nebel (2019) provides this rejoinder rather
forcefully.

12
 Even if the second or the third response succeeds, the traditional is not
out of the woods. It is often overlooked that that-clauses compose with lexical
categories other than verbs.

(13) Now that Andrew had arrived, we smelled durian. adverb

(14) Andrew has the belief that durian is pungent. noun

Such productivity is difficult to explain if that-clauses are designative. Unlike

verbs such as believe that are traditionally thought to require an argument in the
form of a that-clause, the expressions from other lexical categories do not take
arguments. Accordingly, the that-clause is entirely optional. (13) is especially
interesting because it is surprising that now composes with a clause.
A semantics with more promise for explaining the productivity of that-
clauses treats them as predicates. In (14), for example, the clause is a predicate
of the noun belief that specifies the content of the belief. The details differ from
one predicativist semantics to the next.21 To illustrate the semantics owed to
Moulton (2015), which I favor, consider that-clauses with verbs. On his view,
they are not arguments to verbs. Instead, they are a predicate of an argument
slot that other expressions can fill like propositional anaphors. In response, some
might be tempted to argue that the that-clauses found in (13) and (14) have
a different meaning from those found in (7) and (8). The earlier clauses are
referring while the latter clauses are predicates. Without independent evidence
for ambiguity, however, saving referentiality through ambiguity is unmotivated.
I presently know of no such evidence.22

3.2.3 Propositional anaphors

An anaphor is a context-sensitive expression that designates a prominent indi-

vidual. The paradigm example of an anaphor is a pronoun like she, he, or they.
A pronoun has a few different uses besides an anaphoric use, but when it is used
anaphorically it designates an individual introduced earlier in the discourse. We
can observe as much with he in sentence (15).

(15) Andrew had arrived. He smelled of durian.

21
For various proposals, see Kratzer (2006), Arsenijevic (2009), Moulton (2009, 2015, 2017),
and Moltmann (2018).
22
One line taken by predicativists like Arsenijevic (2009) is that all that-clauses are relative
clauses. But de Cuba (2017) notes that various languages use different clause-linking expressions
for relative clauses and that-clauses composing with nouns. However, this does not provide
evidence for the right kind of ambiguity. What is needed is evidence that postnominal and
postadverbial that-clauses are different from postverbal ones.

13
Many natural languages are taken to have anaphors exclusively for propositions.
The most widely accepted example is the expression so.23 Though it has other
uses, it behaves like a designative expression in a postverbal position. The
example below shows how it patterns with it and that.

⎧ it ⎫
(16) (a) Andrew believes durian is pungent.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪ ⎪
(b) I believe ⎨ so ⎬ too.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ that ⎪
⎭
Other expressions regarded as propositional anaphors include the response
markers yes and no.24 Such terms express speaker agreement or disagreement
with a prominent proposition from the discourse. Like so, response markers can
be the argument to a verb.
⎧
⎪ said ⎫ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
(17) I ⎨ think ⎬ yes.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ guess(ed)⎪
⎭
The propositional anaphors distinguish themselves from that-clauses by not be-
ing able to surface in postadverbial or postnominal positions. Ungrammaticality
is produced in the earlier (13) and (14) if the that-clauses are replaced with
anaphors. (18) and (19) show as much.

(18) # Now so, we smelled durian.

(19) # Andrew has the belief so.

See van Elswyk (forthcoming) for a detailed discussion of how substitution

failures between that-clauses and anaphors provide additional reason to maintain
that-clauses are predicative as opposed to designative.
An alternative proposal found in the syntax literature is that response markers
are not anaphoric because their anaphor-like behavior is owed to the fact that
they are obligatorily followed by an elided tense phrase.25 Where the crossed
out text below represents ellipsis, the proposal is that a marker made in response
to (20a) is trailed by elided material.

(20) (a) Durian is pungent.

(b) Yes, durian is pungent.
23
See Kiparsky and Kiparsky (1970), Cushing (1972), Cornish (1992), Needham (2012), Sailor
(2012), and van Elswyk (2019, forthcoming).
24
See Krifka (2013) and van Elswyk (2019).
25
See Kramer and Rawlins (2012), van Craenenbroeck (2010), Holmberg (2013), and Authier
(2013) for this alternative.

14
As van Elswyk (2019) argues, an ellipsis explanation of the response markers
cannot explain their embedding behavior in conditional antecedents or as the
argument to verbs where the non-elided variant is ungrammatical. The balance
of evidence therefore favors an anaphoric explanation on which (A2) is satisfied
because they are designative expressions.

3.2.4 Proposition descriptions

Descriptions like The proposition that durian is pungent have figured in discussions
of (R2) mostly because they fail to be intersubstitutible with that-clauses. But
they are worth considering independently because they plausibly belong to a
broader family of descriptions that include The claim that durian is pungent and
The evidence that durian is pungent and which have been taken to designate propo-
sitions. Let content descriptions name this broader family of descriptions.
Content descriptions presumably have whatever semantics definites have.
But data provided by Nebel (2019, 81) suggests that they are concealed
questions. A concealed question is a determiner phrase that has a question-like
meaning. His evidence involves the verb explains. It only accepts complements
that are determiner phrases if they are interpreted as concealed questions.
⎧
⎪ claim ⎫ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
(21) Andrew explained the ⎨ evidence ⎬ that durian is pungent.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ proposition⎪
⎭
Such data is importantly limited, however. That explains only selects for
determiner phrases that are concealed questions does not mean that such phrases
are always concealed questions. It just means that the content description is a
concealed question in (21). Since determiner phrases that can be interpreted
as concealed questions can be interpreted as normal definites in most other
contexts, content descriptions might still have whatever semantics definites
usually have.
Nebel’s final proposal is that content descriptions are individual concepts.
An individual concept is the intension of an individual, or a function from a
circumstance of evaluation like a possible world to an individual (Montague,
1974). His proposal naturally fits with a view of concealed questions where
they are also individual concepts.26 Descriptions like The proposition that durian
is pungent are constant functions that designate the same proposition in every
world. They contrast with descriptions like The evidence that durian is pungent.
They are variable functions to a proposition that depend on what the evidence
is at the world of evaluation.
26
See Frana (2017) for a recent defense of such a view.

15
 His final proposal will not do, however. Moltmann (2013, 135) notes of
content descriptions that they compose with causal predicates. For example,
caused astonishment is predicated of a claim in (22). However, propositions nor
individual concepts, if abstract, the kinds of entities that can enter into causal
relations. And yet, Nebel’s semantic proposal predicts as much.

(22) The statement that durian isn’t pungent caused astonishment.

The flat-footed understanding of content descriptions is that the designate an

entity of the kind identified by the noun. In (22), for example, the description
designates a statement. A statement is an event, an event in which a speaker uses
a sentence to express a content in a context. That event can be a cause.
One might try the line that content descriptions are ambiguous. But
Moltmann (2013, 137) notes that truth or falsity and causal relations can
simultaneously be predicated of whatever content descriptions designate. Were
they ambiguous, (23) would be strange sounding.

(23) The obviously false statement that durian isn’t pungent caused
astonishment.

The lesson to draw is that proposition descriptions are markedly different from
content descriptions. Whatever content descriptions designate, the latter do not
designate propositions. Accordingly, they fail to satisfy (A3).
However, proposition descriptions should be eliminated from our discussion
of an (R2)-basis for propositions too. Descriptions like The proposition that durian
is pungent are a peculiar bit of philosopher’s English. Native speakers rarely,
if ever, use proposition to identify what philosopher’s mean by the term. It is
typically used to identify a policy or proposal being considered by a decision-
making body. So looking to propositions descriptions for independent evidence
that propositions are needed for (R2) is far from compelling.

3.2.5 Proposition names

Like descriptions, -ism names like logicism or Marxism presumably have whatever
semantics names do. They satisfy (A2) insofar as names do. But, unsurprisingly,
Nebel proposes they are individual concepts. The explain data replicates in that
-ism names can effortlessly be the argument to explain but other names cannot
unless unusual contexts are imagined to enable their interpretation as concealed
questions. However, my earlier remark about the data’s limitation still holds.
Nebel (2019, 93) has another line of data. He notes that (25) is natural in
comparison to (24).

(24) ? Logicism is (# identical to) that arithmetic reduces to logic.

16
 ⎧
⎪ claim ⎫ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
(25) Logicism is (identical to) the ⎨ evidence ⎬ that arithmetic
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ proposition ⎪
⎭
reduces to logic.

His explanation appeals to -ism names being individual concepts. They can fig-
ure in identity statements with content descriptions because content descriptions
are also individual concepts. They cannot figure in identity statements with that-
clauses because the latter are referential. Though Nebel’s explanation assumes
that that-clauses are referential, no help is found if we regard that-clauses as
predicates. A predicative statement like (26) is grammatical.
⎧
⎪ claim ⎫ ⎪
⎪
⎪
⎪ ⎪
⎪
⎪
(26) The ⎨ evidence ⎬ is that arithmetic reduces to logic.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩proposition⎪
⎭
But, as (24) showed us, logicism cannot pattern with descriptions like The claim /
evidence / proposition in (24) to receive a that-clause as a predicate. Tentatively,
then, I endorse the conclusion that -ism names are individual concepts that take
us to a proposition-like entity.

3.3 The case for (A3)

Having considered which entity terms satisfy (A2), let’s consider whether those
terms designate propositions as opposed to other entities. I eliminated that-
clauses and descriptions from the running. What remains are free relatives,
names, and propositional anaphors. I will group names and free relatives
together because they evidence that they designate propositions comes from
the predicates they compose with. In contrast, the evidence that the anaphors
designate proposition mostly comes from the way they are licensed in a context.

3.3.1 Names and free relatives

We can motivate that names and free relatives designate propositions by noticing
that they combine with natural language predicates corresponding to the various
roles enumerated earlier. Consider (R7), for example. Whatever proposition
names and free relatives pick out, (27) shows that it is the kind of entity which
can be believed and known. Or consider the logical roles (R4), (R5), and (R6).
Whatever proposition names and free relatives designate, (28) shows it is also
the kind of entity that can be entailed, true, or possible.

(27) { } is { }.
Logicism believable
What was said known (by Andrew)

17
 ⎧
⎪ ⎫
⎪
⎪
⎪
⎪
true ⎪
⎪
⎪
(28) { } is ⎨ ⎬.
Logicism
⎪
⎪entailed (by logicism)⎪
possible
What was said ⎪
⎪ ⎪
⎪
⎪
⎩ ⎭
Other roles can be captured by adjusting the verb in the free relative. Consider
(R3), which we will discuss independently soon. Relatives like what was guessed
presumably concern the content of illocutionary acts.
Insofar as a proposition is an entity which performs or is suited to perform
(R3) through (R7), the truth of sentences like (27) and (28) evinces that
propositional names and free relatives designate propositions. Let’s call this
the evidence from predicates. To deflate the evidence, two strategies present
themselves. The first is to argue that the adjectives in (27) and (28) are not
predicates. Not wanting to reify truth as a property, many have argued that true
does something else.27 A survey of such proposals is beyond the scope of this
chapter. Suffice it to say that skepticism is an appropriate attitude if such theories
do not consider true’s distribution in a variety of constructions.28 Nevertheless,
the first strategy has not been taken for any of the other adjectives. It is difficult
to see how it could be undertaken without rethinking the canonical relationship
between adjectives and predicates. As matters stand, the first strategy poses little
threat to the evidence from predicates.
The second strategy is to maintain that names and free relatives designate
entities other than propositions and that these other entities can also perform
roles like (R3) through (R7). As I discussed at the start of this section, most taking
this second strategy adopt an easy as opposed to hard non-propositionalism. An
exception is Friederike Moltmann who has prolifically worked out an alternative
with what she calls attitudinal objects, or, for brevity’s sake, what I will call
a-objects. I will briefly consider her alternative view as it is presented in her
2013.
An a-object is what is designated by a descriptions like The belief that
durian is pungent and phrases such as Andrew’s belief that durian is pungent.
When it comes to their metaphysics, a-objects are a special kind of trope or
particularized property. That is, they are quasi-relation tropes that an agent
instantiates and which consist of (a) an attitude mode and (b) the propositional
constituents to which an attitude is to be related. For example, Moltmann
(2013, 158) represents the a-object designated by The belief that Mary likes Bill as
λx.[believe[x; LIKE, Mary, Bill]]. Two a-objects are identical if and only if they
have the same mode and constituents. To illustrate, the a-object designated by
27
See Grover et al. (1975), Brandom (1994), and Moltmann (forthcoming).
28
I have in mind interaction with different subjects (e.g. free relatives, names, quantifiers,
anaphors, sentential subjects), tenses other than the present, epistemic verbs like the seems in
logicism seems true, and adverbs like probably. No such semantics has yet to be undertaken by
anyone offering a non-predicative semantics.

18
The belief that Mary likes Bill is different from the one designated by The hope that
Mary likes Bill because their modes differ.
a-objects differ from propositions because they are not abstract. As tropes,
they are instantiations of properties in space and time. To defend their concrete
status, Moltmann (2013, §4.1.3) notes that all sorts of properties can be
predicated of a-objects which require them to be concrete. Reconsider The
statement that durian is pungent caused astonishment, which was presented earlier
as (22). Insofar as only concreta can enter into causal relations, the sentence’s
felicity shows that the a-object designated by the The statement is concrete as
opposed to abstract. However, a-objects are like propositions in that they are
apt to perform most of the theoretical roles. Or, so Moltmann alleges. Consider
having truth-conditions. That a-objects are true or false is witnessed again by
linguistic evidence. We can form sentences like Andrew’s belief that durian is
pungent is true to predicate truth of an a-object.
Moltmann’s proposal is difficult to assess. We are told that a-objects crucially
involve attitudinal modes as constituents or parts or what-have-you. However,
Moltmann neither explains what such modes are nor how particular modes are
associated with a particular a-object. Attitudinal modes are somehow associated
with the noun in descriptions. But then the specter of circularity surfaces. If
descriptions like The hope that Mary likes Bill designates an a-object with the
attitudinal hope mode, then we travel in a circle if the attitudinal hope mode
is merely the mode that figures in the a-object designated by a description like
The hope that Mary likes Bill.
The linguistic evidence that Moltmann cites to motivate her core claims
about a-objects is also questionable in various ways. For example, Moltmann
(2013, 134) appeals to the infelicity of sentences like (29) as evidence that
attitudinal modes partially determine identity.

(29) ??? John’s claim that it will rain is his hope that it will rain.

But such infelicity strikes me as pragmatic. It is difficult to imagine a context in

which (29) would be uttered. Other sentences that do seem to identify a-objects
with different modes are felicitous.

(30) John claimed what he hoped for—that it will rain.

(31) John expressed his hope through the claim that it will rain.

It is even easier to construct felicitous instances of cross-attitudinal identifications

if the identifications involve different subjects and we use about-phrases to specify
content as opposed to that-clauses.

(32) John’s hope for what will happen is the same as Bill’s belief
about what will happen.

19
 (33) What John permitted is the same as what Bill required.
(34) John’s thought about what should be required is the same as
what Bill commanded.

Since Moltmann bases her individuation claim exclusively on linguistic evidence

like (29), the availability of evidence like (30) through (34) shows that con-
stituents and attitudinal modes are not enough for individuation.
But there is a bigger problem. Such data is very difficult to explain without
propositions. With propositions, it is naturally explained if what is being
identified is the propositional content of attitudes. There is nothing unusual
about saying that the content of John’s hope is the same as the content of Bill’s
belief. On a picture where there are attitudes and propositions at which those
attitudes are directed, two attitudes can differ as relations but be related to the
same proposition. But when we try to do without propositions, matters become
very unusual. The felicity of (30) through (34) appears to saddle Moltmann with
the false conclusion that different attitudes had by different agents are somehow
the same a-object.
Setting aside the general issues with her proposal, it does not facilitate a
non-propositional explanation of -ism names. Though it plausibly extends
to free relatives because they contain a verb like said or guessed that can be
associated with an attitudinal mode, proposition names like logicism do not
involve attitudes. Note too that when we metalingustically introduce -ism names
through stipulations like Let cannot-be-rightism be the view that there are no
propositions or a-objects, we only have an attitude toward the name through the
act of baptism. No attitude by anyone ever needs to be taken to the content that
was named. Attempts to somehow associate every -ism name with a particular
attitudinal mode are slated to fail.
For these reasons, I do not think the second strategy, as taken by Moltmann,
is effective for resisting the evidence from predicates. For both free relatives and
propositional names, but especially for names, the evidence suggests that (A3) is
satisfied: the entity terms designate propositions.
The next bit of evidence is a variant of the evidence from relational attitudes.
It originally aimed to conclude that that-clauses refer to propositions because
attitude verbs are relations between a subject and a proposition and that-clauses
are arguments to attitude verbs. But abandoning the view that that-clauses
designate propositions, which I suggested above in favor of the view that they are
predicates, does not require us to ditch the view that attitude verbs are relations
between (at least) a subject and a proposition as the attitude’s object. Accordingly,
we can find similar evidence with other terms.

(33) Andrew believes { }.

logicism
What was said

20
On the assumption that attitudes like believe relate a subject to a proposition,
which is the linguistic reflex of role (R7), the capacity of propositional names
and free relatives to be the apparent argument to that verb provides a reason
why these terms designate propositions.

3.3.2 Propositional anaphors

Whether with so or the response markers yes and no, the evidence from relational
attitudes can be re-run again for propositional anaphors. The earlier examples
(16) and (17) showed that they can be the arguments to attitudes like believe. As
I discuss elsewhere (van Elswyk, 2019, forthcoming), the reasons for thinking
that-clauses are predicates do not apply to the anaphors. For example, the
anaphors are not intersubstitutible with postnominal that-clauses.

(34) The belief that durian isn’t pungent is false.

⎧
⎪ so ⎫
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪
(35) # The belief ⎨yes⎬ is false.
⎪
⎪
⎪ ⎪
⎪
⎪
⎪
⎩ no ⎪
⎭
Accordingly, the evidence from relational attitude starts us off with a reason why
propositional anaphors satisfy (A3).
But it is not the only reason. The reason given in van Elswyk (2019) depends
on how the anaphors are licensed. Anaphors of any kind cannot be felicitously
used in a context without an antecedent expression introducing an entity of
the proper kind to be available for reference. A sentence like It is pungent is
infelicitous, for example, unless a non-animate entity like durian has recently
been made available for the pronoun it to reference. The propositional anaphors
are typically licensed by tense phrases.29 On the assumption that a proposition is
a representation of an entity being a certain way, tense phrases have a proposition
for their meanings in a context because they are the smallest part of a sentence
in which one finds such a representation of an entity. Let’s call this the evidence
from licensing.
The evidence from licensing, as I have framed it, implicitly answers a concern
from Asher (1993) about using propositional anaphora as evidence for the
existence of propositions. With a variety of examples involving it and that being
used to anaphorically designate propositions, he concludes that the licensing for
it and that being used as propositional anaphors is very messy. Accordingly, were
we to use facts about whether it and that can be used as propositional anaphors
29
See Snider (2017) for defense of the broader claim that they are licensed by operators that
take propositions as arguments. His generalization already assumes that propositions can be
assigned as meanings to sentences or constituents thereof. So it is not useful to our present
purpose of independently defending (A3).

21
as an indicator of whether a proposition is introduced into the conversation,
we would likely conclude, as Asher (1993, 135) does, that the individuation
of propositions is “highly context-sensitive and interest-relative.” From there,
Asher motivates what he calls a conceptualism about propositions. Propositions
come and go as we need them to be designated in discourse, but they otherwise
do not independently exist.
Importantly, though, the licensing of so and the response markers is not
messy. For example, the proposition that is the meaning of tense phrase in
the main clause of a declarative can always be designated by any of them.
There are zero exceptions. The dedicated anaphors, as I called them in van
Elswyk (forthcoming), therefore provide a much stronger basis for (R3) than
the expressions it and that which must be flexibly recruited to be propositional
anaphors.
The most straightforward way to resist the evidence from licensing is to deny
that so and the response markers are anaphors. I discussed earlier the alternative
that they are non-anaphoric terms that are mandatorily followed by an elided
tense phrase. The only other non-propositional alternative that has been
proposed is owed to Lebens (2017). Applying a proposal of Wettstein (2004),
he suggests that propositional anaphors facilitate sentential display. To (re-
)display a sentence, is to present it and its linguistic properties for consideration.
Such a proposal falls into the category of an easy non-propositionalism. A
worked out theory of display is never detailed that explains the relevant meaning
facts surrounding the anaphors. Whatever display amounts to, it will need to
accommodate how anaphors are constrained by the meaning of the tense phrase.
Such an observation becomes important with context-sensitive expressions like
the indexical I.

(36) (a) Andrew thinks that I smell of durian.

(b) Brad thinks so too.

The meaning of (36b) is that Brad also thinks that the speaker of (36a) smells of
durian. The I, in other words, refers to the speaker in (36a) and that contribution
to the meaning of the tense phrase in the embedded clause is what so reproduces.
Such data is easy to explain if the expressions are treated as anaphors for
propositions. The indexical in a context contributes to a proposition and that
proposition is what is available for anaphoric for reference. In contrast, sentential
display might run the risk of merely displayed the embedded clause such that
the indexical I refers to the speaker of (36b) as opposed to (36a).

22
4 The content of illocutionary acts

4.1 Force/content distinction

A speaker’s use of a sentence in a context constitutes an action. To understand

the interior of a speech act, Frege (1918/1956) distinguished between the act’s
force and content. The content of the act is the meaning expressed by the
sentence. The act’s force is what the speaker does with the content.
With regards to speech acts performed by using a declarative, the force/content
is often presented in a manner that assumes (R1). There is the proposition
that is the meaning of a declarative in context. Then there is what speakers
illocutionarily do with that content. However, the distinction does not pre-
suppose that propositions perform (R1). In recent years, numerous arguments
have been given that the meaning of a declarative in a context—its semantic
value—is not a proposition where a proposition is understood to be an entity
like a set of worlds.30 Instead, it is a different entity. These authors therefore
draw a distinction between a declarative’s semantic value and its illocutionary
content. It is because (R1) and (R3) are different theoretical roles with different
requirements for the entities occuping them that the semantic value of a
declarative is thought to not be a proposition. For these authors, to get from
a semantic value to a content is a postsemantic process.31
I do not wade into this particular dispute over the meanings of declaratives.
It is worth flagging, though, not just as another obstacle to an (R1)-basis for
propositions, but also because it helps to illustrate that theoretically (R3) can be
understood in a manner that does not presuppose (R1).
Some deny the force/content distinction in working out a theory of proposi-
tions. Hanks (2007, 2015) is front-and-center here. On the Fregean conception
of content, we can think of predication as a neutral act that is prior to a cognitive
or illocutionary act like judging a predication as true. Hanks argues there is
no neutral act of predication. To predicate is to simultaneously perform a
cognitive or illocutionary act. The force/content distinction therefore needs to
be abandoned to facilitate a theory of propositions as cognitive act types.
I am not convinced. The distinction remains useful for explaining various
facts about communication. To start, I think precisely because there is a neutral
act of predication that is prior to judgment, force and content need to be
distinguished. For a response to Hank along these lines, see Green (2018).
Additionally, the force/content distinction is useful for explaining how different
30
See Stanley (1997), Ninan (2010, 2012), Rabern (2012), and Yli-Vakkuri (2013). Earlier
versions of this argument are found in Dummett (1973) and Lewis (1980).
31
See Yalcin (2014), Stojnic (2016), and Rabern (2017) for recent discussion of such
postsemantic processes.

23
speech acts can have the same content. Frege (1918/1956) initially made this
point with declaratives and polar interrogatives like the pair below.

(37) Durian is pungent.

(38) Is durian pungent?

For him, they had the same content but different illocutionary force. I think
a nearby point holds true. Since I take a proposition to be the meaning of a
tense phrase in a context, both (37) and (38) contain the same tense phrase.
Supporting evidence comes from propositional anaphors. Consider how Yes, it
is or I think so are both felicitous responses to (37) and (38).
But the point can be without the assumption about the meaning of tense
phrases. Recall my earlier examples in §3.3.1 of cross-attitudinal identifica-
tions. We can tweak these examples to be about speech act reports by using
corresponding verbs like assert and guess.

(39) Andrew asserted what Brad merely guessed—that durian is

pungent.

The natural interpretation of (39) is that it identifies the content of two speech
acts with difference force. If force and content are not distinguished, it is
difficult to understand how sentences like (39) could be true let alone what they
could mean. Finally, distinguishing force and content allow the two aspects
to be explained differently. Content is determined by semantic convention
plus a postsemantic process or two. The distinction allows force to be non-
conventional. It can be produced by speaker intentions, social norms, con-
stitutive rules, and more. Without the distinction, however, force must be
conventional. Though I myself take the minority view that much of what we
think of as force is semantic (van Elswyk, 2018), it cannot all be. For example,
the social practices of holding speakers liable to blame or censure for insincerity
is traditionally regarded as an aspect of force and that aspect is not owed to
semantic conventions.

4.2 Assertion

As tradition has it, there are many distinct speech acts which can be performed by
using a declarative in a context. I focus on the speech act of assertion in particular
because it is the typical speech act performed with a declarative. Theories of
assertion come in many different varieties.32 Following Benton and van Elswyk
32
See Cappelen (2011) and MacFarlane (2011) for recent taxonomies.

24
(2018), I distinguish theories between those that are representational and non-
representational. A representational theory, at least in part, characterizes asser-
tion as an act in which the speaker expresses or represents her epistemic position
towards illocutionary content. In contrast, non-representational theories do not
characterize assertion with reference to the speaker’s epistemic position.
Many different theories fit the description of representational. I highlight
two. The first is a norm-based theory of the kind popularized by Williamson
(2000). On this approach, assertion is characterized by a norm requiring the
fulfillment of an epistemic condition C.

norm-based theory schema

One must: assert that p only if C.

Candidate conditions include knowledge, justified belief, and truth. The second
approach is an effects-based theory of the kind popularized by Stalnaker (1978).
This approach characterizes assertion according to what it does. For Stalnaker,
the characteristic effect of assertion is a proposal to update the common ground
of a context. The common ground is the set of propositions that the conversa-
tional participants are mutually believing or accepting in the context.
Both of these accounts straightforwardly assume the existence of proposi-
tions. But we can abstract away from this assumption to focus more generally
on the requirements they place on illocutionary content. For both, illocutionary
content is something towards which a speaker has a cognitive attitude. On a
norm-based theory, to assert is to be governed by a norm requiring the speaker
to take an attitude like knowledge or belief to illocutionary content. Likewise, to
assert is to change what participants take an attitude like belief or acceptance
towards. We therefore have good reason to link (R3) with (R7). Given a
representational theory, whatever is the object of cognitive attitudes is what the
content of assertion is.
It is not a stretch to conclude from the linking of (R3) and (R7) that a propo-
sition is what fulfills both because a proposition, on a functional characterization,
is what does or is apt to perform the roles. I dub this the evidence from assertion.
However, there are at two familiar ways to counter that it is a stretch. The first
is to plumb for a non-representational theory of assertion that does not thereby
require us to link (R3) with (R7). I do not regard non-representational theories
of assertion as being as explanatory and have argued as much elsewhere.33
The second counter is to argue that another entity plays the linked roles or
there is a way to proceed with them. Lebens (2017) takes up this challenge
by reviving a multiple relations theory of judgment (mrtj) initially proposed
by Russell (1910). On an mrtj, judgment is not an unary relation between a
33
See Benton and van Elswyk (2018) and van Elswyk (2018).

25
subject a proposition. It is a polyadic relation to what the proposition represents.
Where a propositionalist would say that the judgment that durian is pungent
is judgment towards a proposition, Lebens says that we are related to durian
and pungentness such that the latter is judged to be the former. Lebens
thereby explains an attitudinal notion like the common ground in terms of
being similarly judging. Where a propositionalist would say that a proposition
enters the common ground if accepted, Lebens says grounding happens when
conversational participants both stand in the acceptance relation, a kind of
judgment, to various objects and properties ordered in the same way.
The mrtj or something like it does not pose a viable alternative. Beyond
the broader worry that Lebens’s revival of mrtj does not overcome the initial
problems that plagued Russell’s version, it obliterates the force/content distinc-
tion. Lebens takes judgments to be the bearers of truth and falsity as opposed
to contents to which judgments are related. As subkinds of judgment, speech
acts are therefore true or false. Given mrtj, there cannot then be speech acts like
assertions and guesses that differ in force but share the same content.

5 Conclusion

For those who deny their existence, propositions are creatures of darkness
only countenanced because theorizing got carried way (Quine, 1960). Others
may admit their existence, but worry, as Lewis (1986, 54) aptly does, that
“the conception we associate with the word proposition may be something of a
jumble of conflicting desiderata.” Whether such reactions are justified depends
on whether there are roles for propositions to play in our leading theories.
This chapter has surveyed some of the details pertinent to whether propo-
sitions are needed for the linguistic roles. Given the theoretical choice-points
we have considered, the linguistic basis for propositions remains. But not as
the basis is usually championed. When it comes to (R1), positing propositions
is not necessary. Other entities can be and are assigned as the meanings of
declaratives in leading semantic theories. But the propositionalist receives a
consolation prize. Though other entities can and do supplant propositions in
successful theories, theories often retain one or more theoretical roles for a
proposition. When it comes to (R2), I have suggested that propositions are
plausibly needed. They are needed to explain what certain expressions of natural
language designate. Importantly, though, that-clauses do not make this list. The
list includes propositional anaphors, free relatives, and names. Finally, when
it comes to (R3), I have more tentatively suggested that leading theories of
assertion require illocutionary content to be a proposition.
My own view on the place of propositions fits well with this conclusion. I

26
take propositions to be the meanings of tense phrases in a context.34 Whether
or not propositions fulfill (R1), propositions therefore perform the auxiliary role
(R1**). Since tense phrases are what license the use of propositional anaphors,
(R1**) and (R2) are tightly connected. The availability for propositional
anaphora is a necessary condition for whether content is assertoric or not. In
other words, there is no situation in which th object of assertion cannot be
targeted with anaphors. So there is a tight connection between (R1**), (R2),
and (R3) as well.
However, the connectedness of the roles is not mandatory. Of particular
independence from the rest is (R1). Independence can be illustrated by
imagining different ways to reject that propositions are needed to perform (R1)
while accepting that they are still needed for some role. One might adopt a
truth-thereotic semantics and thereby not assign propositions as the meaning
of any declaratives or phrases within one. And yet, propositions might still be
countenanced to explain what propositional names and anaphors designate in
a context. Or, one might regard the semantic value of a declarative as being a
different entity entirely, but still appeal to a postsemantic process to convert that
entity into a proposition to be assertoric content. Accordingly, there are many
ways to break with tradition or my opinionated survey and still find cause to posit
propositions.35

34
Taking propositions to be the meanings of tense phrases is a familiar enough assumption in
semantic theories. However, independent evidence comes from the distinction between tense
phrases and small clauses, language acquisition, and propositional anaphora. See Glanzberg
(2011), Roeper (2011), and van Elswyk (2019) for discussion. 35 Thanks are owed to Andrew
Bailey and Joshua Spencer for comments or conversation.

27
References

Armour-Garb, Bradley and Woodbridge, Cinque, Guglielmo. 1999. Adverbs and

James. 2012. “The Story About Proposi- Functional Heads: A Crosslinguistic Perspective.
tions.” Nous 46:635–674. Oxford: Oxford University Press.
Arsenijevic, Boban. 2009. “Clausal comple- Colyvan, Mark. 2010. “There is No Easy
mentation as relativization.” Lingua 119:39– Road to Nominalism.” Mind 119:285–306.
50. Cornish, Francis. 1992. “So Be It: The
Asher, Nicholas. 1993. Reference to Abstract Discourse Semantic Roles of So and It.”
Objects in Discourse. Springer. Journal of Semantics 9:163–178.

Authier, J.-Marc. 2013. “Phase-Edge Fea- Crawford, Sean. 2014. “Propositional or

tures and the Syntax of Polarity Particles.” non-propositional attitudes?” Philosophical
Linguistic Inquiry 44:345–389. Studies 168:179–210.
Cushing, Steven. 1972. “The Semantics of
Bach, Kent. 1997. “Do belief reports
Sentence Pronominalization.” Foundations of
report beliefs?” Pacific Philosophical Quarterly
Language 9:186–208.
79:215–241.
Davidson, Donald. 1967. “Truth and
Ball, Derek and Rabern, Brian. 2018. “Intro-
meaning.” Synthese 17:304–323.
duction to the Science of Meaning.” In Derek
Ball and Brian Rabern (eds.), The science of Dayal, Veneeta. 1997. “Free relatives and
meaning. Oxford University Press. ever: Identity and free choice readings.” In
Aaron Lawson (ed.), SALT 7, 99–116. CLC
Bealer, George. 1998. “Propositions.” Mind Publications.
107:1–32.
de Cuba, Carlos. 2017. “Noun complement
Benton, Mathew and van Elswyk, Peter. clauses as referential modifiers.” Glossa 3:1–
2018. “Hedged Assertion.” In Sandy 46.
Goldberg (ed.), Oxford Handbook of Assertion.
Oxford University Press. Dolby, David. 2009. “The Reference
Principle: A Defence.” Analysis 69:286–296.
Berman, Stephen. 1991. On the semantics
Dummett, Michael. 1973. Frege: Philosophy
and logical form of wh-clauses. Ph.D. thesis,
of Language. Duckworth.
University of Massachusetts, Amherst.
Frana, Ilaria. 2017. Concealed Questions.
Brandom, Robert. 1994. Making It Explicit:
Oxford University Press.
Reasoning, Representing, and Discursive Com-
mitment. Harvard University Press. Frege, Gottlob. 1918/1956. “The Thought:
a logical inquiry.” Mind 65:289–311.
Burge, Tyler. 1979. “Individualism and the
mental.” Midwest Studies in Philosophy 4:73– Glanzberg, Michael. 2011. “More on
121. operators and tense.” Analysis 71:112–123.
—. 2014. “Partiality and Explanation in
Cappelen, Herman. 2011. “Against As-
Semantics.” In A. Burgress and B. Sherman
sertion.” In Jessica Brown and Herman
(eds.), Metasemantics, 259–292. Oxford Uni-
Cappelen (eds.), Assertion. Oxford: Oxford
versity Press.
University Press.
Green, Mitchell. 2018. “A refinement and
Chomsky, Noam. 1995. “Language and
Defense of the Force/Content Distinction.”
Nature.” Mind 104:1–61.
In Daniel Fogal, Daniel Harris, and Matt
—. 1996. Powers and Prospects. Boston: South Matt (eds.), New Work On Speech Acts. Oxford
End Press. University Press.

28
Grover, Dorothy, Kamp, Joseph, and Belnap, Kramer, Ruth and Rawlins, Kyle. 2012.
Nuel. 1975. “A Prosentential Theory of “Polarity particles: an ellipsis account.” In
Truth.” Philosophical Studies 27:73–125. Suzi Lima, Kevin Mullin, and Brian Smith
(eds.), Proceedings of NELS 39.
Hale, Bob. 1994. “Singular terms.” In
B. McGuinness and G. Oliveri (eds.), The Kratzer, Angelika. 2006.
philosophy of Michael Dummett, 17–44. Dor- “Decomposing attitude verbs.”
drecht: Kluwer. Http://semanticsarchive.net/Archive/DcwY2JkM/attitude-
verbs2006.pdf.
—. 1996. “Singular terms.” In M. Schirn
(ed.), Frege: Importance and legacy, 438–457. Krifka, Manfred. 2013. “Response particles
De Gruyter. as propositional anaphors.” In Todd Snider
(ed.), Proceedings of SALT, volume 23, 1–18.
Hanks, Peter. 2007. “The Content-Force
—. 2015. “Bias in Commitment Space
Distinction.” Philosophical Studies 134:141–
Semantics: Declarative questions, negated
164.
questions, and question tags.” Semantics and
—. 2015. Propositional Content. Oxford: Linguistics Theory 25:328–345.
Oxford University Press.
Lebens, Samuel. 2017. Bertrand Russell and
Hofweber, Thomas. 2007. “Innocent the Nature of Propositions: A History and Defence
statements and their metaphysically loaded of the Multiple Relation Theory of Judgment.
counterparts.” Philosophers’ Imprint 7:1–33. Routledge.

Holmberg, Anders. 2013. “The Syntax of Lepore, Ernest. 1983. “What model–
Answers to Polar Questions in English and theoretic semantics cannot do.” Synthese
Swedish.” Lingua 128:31–50. 54:167–187.
Lepore, Ernest and Ludwig, Kirk. 2007.
Horwich, Paul. 1998. Meaning. Oxford
Donald Davidson’s Truth-Theoretic Semantics.
University Press.
Oxford University Press.
Jacobson, Pauline. 1995. “On the quantifica-
Lewis, David. 1980. “Index, Context and
tional force of English free relatives.” In Em-
Content.” In S. Kanger and S. Ohman (eds.),
mon Bach, Eloise Jelinek, Angelika Kratzer,
Philosophy and Grammar, 79–100. Riedel.
and Barbara Partee (eds.), Quantification in
natural languages, volume II. Kluwer. —. 1986. On the Plurality of Worlds.
Blackwell.
Jubien, Michael. 2001. “Propositions and
Lohndal, Terje and Pietroski, Paul. 2011.
the objects of thought.” Philosophical Studies
“Interrogatives, instructions, and I-languages:
104:47–62.
An I-semantics for questions.” Linguistics
Kartunnen, Lauri and Peters, Stanley. 1979. Analysis 37:459–510.
“Conventional implicature.” In Syntax and
MacFarlane, John. 2011. “What is As-
Semantics, volume 11, 1–56. New York:
sertion?” In Jessica Brown and Herman
Academic Press.
Cappelen (eds.), Assertion. Oxford University
King, Jeffrey C. 2002. “Designating Propo- Press.
sitions.” Philosophical Review 111:341–371. Mandelkern, Matthew. 2019. “Bounded
—. 2007. The Nature and Structure of Content. modality.” Philosophical Review 128:1–61.
Oxford University Press. McCready, Elin. 2010. “Varieties of Conven-
Kiparsky, Paul and Kiparsky, Carol. 1970. tional Implicature.” Semantics and Pragmatics
“Fact.” In M. Bierwisch and K.E. Heidolph 3:1–57.
(eds.), Progress in Linguistics: A Collection of Merricks, Trenton. 2015. Propositions.
Papers, 143–73. Mouton de Gruyter. Oxford: Oxford University Press.

29
Moltmann, Frederike. 2003. “Propositional —. 2012. “Propositions, Semantic Values,
attitudes without propositions.” Synthese and Rigidity.” Philosophical Studies 158:401–
135:70–118. 413.
—. 2013. Abstract Objects and Semantics of Pickel, Bryan. forthcoming. “Structure
Natural Language. Oxford: Oxford University propositions in a generative grammar.” Mind
Press. .
Moltmann, Friederike. 2018. “Clauses as Pietroski, Paul. 2018. Conjoined meanings.
semantic predicates: difficulties for possible- Oxford University Press.
worlds semantics.” In R. Bhatt, I. Frana,
Pollard, Carl. 2015. “Agnostic Hyperinten-
and P. Menendez-Benito (eds.), Festschrift for
sional Semantics.” Synthese 192:535–562.
Angelika Kratzer. University of Massachusetts.
Potts, Christopher. 2004. The Logic of
—. forthcoming. “Truth predicates, truth
Conventional Implicatures. Oxford University
bearers, and their variants.” Synthese .
Press.
Montague, Richard. 1974. Formal Philosophy:
Putnam, Hilary. 1975. “The Meaning
Selected Papers of Richard Montague. New
of Meaning.” In Mind, Language, and
Haven: Yale University Press.
Reality, 215–271. Cambridge: Cambridge
Moss, Sarah. 2018. Probabilistic Knowledge. University Press.
Oxford University Press.
Quine, William Van Orman. 1960. Word and
Moulton, Keir. 2009. Natural Selection and Object. MIT Press.
the Syntax of Clausal Complementation. Ph.D.
Rabern, Brian. 2012. “Against the Identifi-
thesis, University of Masschusetts, Amherst.
cation of Assertoric Content with Composi-
—. 2015. “CPs: copies and compositional- tional Value.” Synthese 189:75–96.
ity.” Linguistic Inquiry 46:305–342.
—. 2017. “A bridge from semantic value to
—. 2017. “Determiners on clauses.” In content.” Philosophical Topics 45:201–226.
N. LaCara, K. Moulton, and A-M. Tessier
Roeper, Tom. 2011. “How the Emergence of
(eds.), A Schrift to Fest Kyle Johnson, 293–304.
Propositions Separates Strict Interfaces from
Linguistics Open Access Publications 1.
General Inference.” Proceedings of Sinn and
Murray, Sarah and Starr, Will. 2018. “Force Bedeutung 15:1–21.
and Conversational States.” In Daniel Fogal,
Rosefeldt, Tobias. 2008. “‘That’-clauses and
Daniel Harris, and Matt Moss (eds.), New
non-nominal quantification.” Philosophical
Work On Speech Acts. Oxford University Press.
Studies 137:301–333.
Muskens, Reinhard. 2005. “Sense and the
Rothschild, Daniel. 2013. “Do Indicative
Computation of Reference.” Linguistics and
Conditionals express propositions?” Nous
Philosophy 28:473–504.
47:49–68.
Nebel, Jacob. 2019. “Hopes, Fears, and Other
Russell, Bertrand. 1910. “On the nature of
Grammatical Scarecrows.” Philosophical Re-
Truth and Falsehood.” In Philosophical Essays,
view 128:63–105.
chapter 147–161. London: Routledge.
Needham, Stephanie. 2012. Propositional
Sailor, Craig. 2012. “You can’t regret so (even
Anaphora in English: The relationship between
if you think so).” Manuscript.
so and discourse. Master’s thesis, Carleton
University. Schiffer, Stephen. 1972. Meaning. Oxford:
Clarendon Press.
Ninan, Dilip. 2010. “Semantics and the Ob-
jects of Assertion.” Linguistics and Philosophy —. 2003. The Things We Mean. Oxford
33:355–380. University Press.

30
Simchen, Ori. forthcoming. “Instrumen- van Elswyk, Peter. 2018. Un/qualified
talism about Structured Propositions.” In declaratives. Ph.D. thesis, Rutgers University,
Chris Tillmann (ed.), Routledge Handbook of New Brunswick.
Propositions. Routledge.
—. 2019. “Propositional anaphors.” Philo-
Snider, Todd. 2017. Anaphoric Reference to sophical Studies 176:1055–1075.
Propositions. Ph.D. thesis, Cornell University.
—. forthcoming. “That-clauses and proposi-
Stalnaker, Robert. 1978. “Assertion.” In tional anaphors.” Philosophical Studies .
Syntax and Semantics 9: Pragmatics, 315–332.
Academic Press. Šimík, Radek. forthcoming. “Free rela-
tives.” In Daniel Gutzmann, Lisa Matthew-
—. 1987. “Semantics for Belief.” Philosophi-
son, Cécile Meier, Hotze Rullmann, and
cal Topics 15:177–190.
Thomas Ede Zimmerman (eds.), Semantics
Stanley, Jason. 1997. “Rigidity and Content.” Companion. Wiley–Blackwell.
In R. Heck (ed.), Language, Thought, and
Logic: Essays in honour of Michael Dummett. Wettstein, Howard. 2004. The Magic Prism:
Oxford: Oxford University Press. An Essay in the Philosophy of Language. Oxford
University Press.
Starr, William B. 2014. “Mood, Force, and
Truth.” ProtoSociology 31:160–181. Williamson, Timothy. 2000. Knowledge and
Its Limits. Oxford University Press.
Stojnic, Una. 2016. “On the Connection
between Semantic Content and the Objects Wright, Crispin. 1998. “Why Frege Did
of Assertion.” Philosophical Topics . Not Deserve His ’Granum Salis.’ A Note On
—. forthcoming. “Content in a Dynamic the Paradox of ‘The Concept Horse’ and
Context.” Nous . the Ascription of ‘Bedeutung’ to Predicates.”
Grazer Philosophische Studien 55:239–263.
Stokke, Andreas. 2014. “Truth and context
change.” Journal of Philosophical Logic 43:33– Yalcin, Seth. 2007. “Epistemic Modals.”
51. Mind 116:983–1026.
Thomason, Richmond. 1980. “A Model —. 2014. “Semantics and Metasemantics in
Theory for Propositional Attitudes.” Lin- the Context of Generative Grammar.” In
gusitics and Philosophy 4:47–70. A. Burgress and B. Sherman (eds.), Metase-
Trueman, Robert. 2012. “Dolby Substitu- mantics, 17–54. Oxford University Press.
tion (Where Available).” Analysis 72:98–102. Yli-Vakkuri, Juhani. 2013. “Propositions and
—. 2018. “Substitution in a sense.” Philosoph- Compositionality.” Philosophical Perspectives
ical Studies 175:3069–3098. 27:526–563.

van Craenenbroeck, Jeroen. 2010. The Yli-Vakkuri, Juhani and Hawthorne, John.
Syntax of Ellipsis: Evidence from Dutch Dialects. 2018. Narrow content. Oxford University
Oxford University Press. Press.

31