Professional Documents
Culture Documents
JUSTIN D’AMBROSIO
Australian National University
Marc Antony
Julius Caesar, act III, scene II
Ted Nugent
1
2 · Justin D’Ambrosio
1. Introduction
Language is a tool of influence. Indeed, the history of political oratory shows that
there is no better tool than language for its exertion—it is immensely powerful.
Marc Antony’s speech turns Caesar’s funeral crowd into a frenzied mob; Donald
Trump tweets and the stock market crashes. But the influence exerted through
such speech is rarely innocent. Marc Antony does not induce a frenzy through
rational persuasion. Trump does not incite crowds and terrify day-traders by
arguing from premises to conclusions. Marc Antony and Trump—like countless
others in both public discourse and private conversation—speak manipulatively,
and it is exactly because their speech is manipulative that it has the power it
does. Manipulative speech has the capacity to subvert our reason, undermine
our agency, and work against our interests, and is standardly taken to be a core
component of propaganda, which has the potential to do all three of these things
on a massive scale.1
But it’s a striking fact that while we know that so much of the speech we en-
counter is manipulative, and we know the harms that manipulative speech can
have, the philosophy of language has no general account of what manipulative
speech is, how it functions, or the variety of forms that it can take. For the most
part, the philosophy of language has focused on speech that is innocent. As
we will see below, the standard account of communication in the philosophy of
language—inspired by Grice—assumes that speech is part of a cooperative enter-
prise, and that speakers aim to make public how their speech acts contribute to
the common goal of conversation. As a consequence, the philosophy of language
has largely overlooked manipulative speech as an important category of inves-
tigation. And even when it has focused on manipulative speech, it has focused
only on particular manipulative speech acts, leaving what such speech acts have
in common—i.e. their manipulative element—unexplained.
This paper develops a theory of manipulative speech. The first component is
a definition of manipulative speech formulated in terms of a speaker’s intentions.
On my view, manipulative speech involves a deliberate, coordinated violation
of the two core Gricean norms for conversation: Cooperativity and Publicity.
Speakers often have goals that conflict with the goals that their interlocutors
have for a conversation. An audience may have the goal of gaining knowledge
concerning a topic of discussion, but a speaker may have reason not to cooperate
with this goal. In such cases, it will often be rational for the speaker to violate
the norm of Cooperativity—it will be rational for them to speak strategically. But
1 Foraccounts of propaganda’s achievements and effects, see Bernays [1928], Herman and
Chomsky [1988], Chomsky [1991], among many others. Nearly everyone who has attempted
to define propaganda agrees that it is closely related to manipulative speech. I discuss this
connection in §6.
A Theory of Manipulative Speech · 3
like “desires”, “things a person takes to be in their interest”, or “goals”. I make this choice for
two reasons. First, construing “interests” in this way allows for the possiblity of paternalistic
manipulation. I take paternalistic manipulation to be manipulation that attempts to influence
someone in ways that are in their interests, but against what they take their interests to be, or at
least against what they immediately want. Second, this usage is consonant with the literature
in decision and game theory, which generally treats utilities in this narrow or internal sense. I
return to this point in §3.
3 In addition to Raz, Noggle [1996], Barnhill [2014], Hanna [2015], Stanley [2015], and
Sunstein [2016] all hold the view that manipulation is a form of influence that perverts or
partially subverts rational deliberation.
4 But there is disagreement over whether manipulation must be covert. A signficant minority
of theorists—including Kligman and Culver [1992], Noggle [1996], Barnhill [2014], and Hanna
[2015]—hold that manipulation need not be hidden, and may operate out in the open. I return
to this point at length in §4.
A Theory of Manipulative Speech · 5
are normative insofar as they specify which contributions to conversation are felicitous, or
as Szabó puts it, they spell out what it is for a conversation to go smoothly. But they are also
descriptive: Grice clearly thinks that conversation is standardly or by default cooperative,
which is why we are justified in presuming Cooperativity when we interpret speakers.
6 · Justin D’Ambrosio
2.1. Cooperativity
Grice phrases his Cooperative Principle as follows:
Grice then goes on to present various “maxims” that spell out what it is to abide
by the cooperative principle: Quantity, Quality, Relation, and Manner.
Quantity :
Relation :
Be relevant.
Manner : Be perspicuous.
These maxims specify what can be expected from cooperative speakers, and in
so doing, serve as a guide for interpreters in trying to calculate what a speaker
means by a particular utterance. When what a speaker says obviously violates
one of these maxims, the Gricean picture does not invite us to conclude that the
speaker is in fact being uncooperative. The Gricean model presumes that speak-
ers are being cooperative, and uses this presumption as a premise in reasoning
A Theory of Manipulative Speech · 7
to the conclusion that the speaker must have meant something other than what
they said.
In spelling out what it is for a speaker to be cooperative, these maxims appear
to place a collection of constraints on contributions to a conversation that must
be balanced against each other. Thought of in this way, the maxims, and so the
cooperative principle, appear to specify what it is for a participant to make an
optimal contribution to a conversation with a particular goal. With this idea in
mind, we can formulate the principle of Cooperativity as follows:
Cooperativity: Ensure that your speech acts in the conversation contribute opti-
mally to the goal of the conversation.
This way of thinking about Grice’s principle is intuitive, has a long history, and
is part of a well-developed program within pragmatics.7 But admittedly, it is
controversial.8 I adopt this formulation for definiteness, and because it has the
advantage of allowing for degrees of cooperativity. But the view of manipulative
speech I present below can be expressed using a variety of alternative formula-
tions.
2.2. Publicity
The principle of Publicity is not one that Grice states explicitly. However, for
the recovery of a speaker’s intentions to be possible, Publicity must be in effect.
Following Szabó [2020], we can state the principle as follows:
Publicity: Ensure that it is common knowledge what speech acts are performed
in the conversation and what the goal of the conversation is. [Szabó, 2020]9
7 The idea that the cooperative principle and its associated maxims are best understood
as optimality constraints is developed at length by Blutner [2000], Blutner and Zeevat [2004],
Blutner et al. [2006], and contrasted with the game-theoretic approach by Franke [2009, Ch.
4]. The OT approach to pragmatics builds on Horn’s [1984] reduction of the Gricean maxims
to his principles Q and R, which together can be seen as specifying the notion of an optimal
contribution to a conversation. Similar principles—the Q and I principles—are discussed by
Levinson [1983].
8 There are many alternative formulations of the principle corresponding to different frame-
works that incorporate Grice’s basic insight. Szabó [2020] phrases Cooperativity as follows:
“Ensure that it is common knowledge how speech acts in the conversation contribute to the
goal of conversation.” Asher and Lascarides [2003, 2013] state what they call the rule of Strong
Cooperativity: “Normally, if A M-intends that φ, then B should intend that φ.” Game-theoretic
accounts of the Gricean program implement cooperativity by treating the sender and receiver
as having aligned preferences (see, for instance, Franke [2009, p. 126]).
9 We might also add the requirement that speakers make it common knowledge how their
speech acts contribute to the goal of the conversation, but for now I will set this complication
to one side.
8 · Justin D’Ambrosio
The reason that Grice must endorse the principle of Publicity is that in order
to engage in pragmatic reasoning, interlocutors need to be able to determine
whether a speech act complies with Cooperativity. If Publicity is not satisfied,
participants cannot determine this. Thus, Publicity is entailed by the very pos-
sibility of calculating an implicature. However, as with Cooperativity, there is
controversy over how to best formulate Publicity, and I adopt this formulation
only for definiteness. The account of manipulative speech I present below can be
recast using a variety of different formulations.10
Cooperativity and Publicity can both be seen as norms that govern conver-
sational participants insofar as those participants have a common conversational
goal. If they do have such a goal, then Cooperativity simply exhorts interlocu-
tors to pursue that goal optimally. Further, if interlocutors know that they have a
common goal, then they will be justified in presuming Cooperativity when they
interpret one another. This reasoning likewise extends to Publicity. If partici-
pants have a common goal, then it will be rational for them to make clear what
that goal is, and how their contributions to the conversation promote it. If one
does not make public which speech act one is engaged in, and what one’s goal
is in undertaking that act, then one’s conversational partner will not reliably be
able to determine what is meant, and so the goal of the conversation will not be
achieved.
mind, and work in pragmatics and the philosophy of language has largely fo-
cused on linguistic exchanges that have this goal.11 Putting this goal together
with the above principles yields the Gricean picture of conversation: conversa-
tion is a joint enterprise in which interlocutors aim to optimally and publically
contribute to the goal of sharing knowledge pertaining to a topic of common
concern.
The idea that the goal of conversation is to share knowledge is a core com-
ponent of a prominent development of the Gricean program due to Robert Stal-
naker.12 On the Stalnakerian model, conversational participants begin conversa-
tion with a store of common knowledge, represented by a set of possible worlds.
Topics of conversation are represented by questions under discussion, which par-
tition that set into cells—the possible answers to the question.13 Cooperativity
can then be formulated as the requirement that speech acts in a conversation con-
tribute (sufficiently, adequately, or perhaps optimally) to answering the question
under discussion by ruling out potential answers. Publicity can be defined as the
requirement that speakers make it common knowledge how each of their speech
acts works toward this goal. Conversation then proceeds sequentially, with inter-
locutors taking turns sharing knowledge, until the store of common knowledge
entails a complete answer to the question under discussion.
But the idea that the goal of conversation is to share knowledge is idealized
in two ways that will become important as our discussion proceeds. First, while
it may be the case that sharing knowledge is standardly the goal of conversation,
even Grice concedes that conversation will often not have this goal. Instead, the
goal of our conversation may be for us to amuse or express ourselves, for me
to direct you in some task, for us to bond, or perhaps for us to form the basis
for future cooperative endeavors. Second, and even more importantly for our
purposes, the assumption that conversations have a single goal, and that this
goal is shared among participants, excludes conversations in which participants
have goals that conflict—it excludes speech that is not fully cooperative from the
scope of the Gricean theory. Developing a theory of manipulative speech requires
us to consider conversations in which this this assumption does not hold, and it
is to such conversations that we now turn.
11 See the first pages of Grice [1989a], Stalnaker [2014], and Szabó [2020]. The idea that
the goal of conversation is to share knowledge is built into many pragmatic frameworks and
models. For criticisms of this view see Beaver and Stanley [2019].
12 See Stalnaker [1998, 2002, 2014]. Stalnaker’s model of conversation treats it as a sequential,
Aragonès and Postlewaite [2002], Dickson and Scheve [2006] and Tomz and van Houweling
[2009] for representative examples.
16 For instance, Lee and Pinker [2010], Fricker [2012], Asher and Lascarides [2013], Asher
et al. [2017], Asher and Paul [2018], Camp [2018], Langton [2018], Beaver and Stanley [2019],
Cappelen and Dever [2019], McGowan [2019], and Quaranto and Stanley [forthcoming]. Work
on slurs and derogatory speech also plausibly falls into the category of strategic speech—see
Bolinger [2017], Bach [2018], and Nunberg [2018] for discussion.
17 See Franke and van Rooij [2015], Henderson and McCready [2018], Saul [2018], and
Stanley [2015] for discussions that touch on manipulative speech. As we will see below, on my
definition, lying and misleading also qualify as manipulative. See Fallis [2009], Saul [2012], and
Stokke [2013] for an overview.
A Theory of Manipulative Speech · 11
In this case, the goal of the prosecutor—and indeed of most people in the court,
including the judge and jury—is to come to know whether Mr. Bronston has,
or has ever had, a Swiss bank account. Mr. Bronston has a goal—the goal of
avoiding jail time—that conflicts with the prosecutor’s goal, so he speaks strate-
gically: he says something that does not directly answer the question. Of course,
in doing so, Bronston also implicates a negative answer to the question, and so
attempts to mislead the prosecutor. We will discuss this case further below.
In each of these cases, a speaker knowingly speaks in a way that conflicts
with the audience’s goal of coming to have knowledge concerning a mutually
salient topic, and in doing so, deliberately speaks in a way that is not fully co-
operative. But such speech does not fit into the Gricean account of conversation
outlined above. The Gricean approach to conversation assumes that conversa-
tions have a single, shared goal—typically understood to be the goal of sharing
knowledge—and assumes that interlocutors are cooperative with respect to this
goal. But in strategic speech, speaker and audience have distinct, conflicting
goals for the same conversation.19 Without specifying how the idea of the goal of
conversation extends to cases in which conversational goals conflict, the Gricean
cannot even say that strategic speech violates Cooperativity, since the principle
of Cooperativity is defined in terms of the goal of conversation.
In order to account for divergent conversational goals, I propose that we
revise the Gricean conception of the goal of the conversation as follows:
(1) Let the salient goal of the conversation be the goal G such that
a. the audience’s goal for the conversation is G, and
b. the speaker knows that the audience’s goal for the conversation is G.
This revision is natural, given the dynamics of the cases just discussed, and it al-
lows us to dispense with the two idealizations introduced by the assumption that
the goal of conversation is to share knowledge. First, the idea that a conversation
has a salient goal does not entail that all participants in the conversation share
that goal. As a consequence, this approach accommodates the fact that in each
of the above cases, the speaker does not share what he knows is the audience’s
19 HereI am making two simplifying assumptions that I will later discharge. First, I am
assuming that audiences are unified or monolithic, so that they can be seen as having a single
goal for the conversation, as opposed to multiple conflicting goals. Second, I am assuming that
all strategic speech occurs as part of conversation. I will discharge both of these assumptions in
§6, when I discuss how my account can be extended to yield an account of propaganda.
A Theory of Manipulative Speech · 13
goal for the conversation.20 Second, the salient goal of the conversation need not
have anything to do with sharing knowledge. While in each of the cases we have
discussed, the salient goal is to share knowledge, the salient goal could just as
easily have been something altogether different.21
With the notion of a conversation’s salient goal in hand, the principles of
Cooperativity and Publicity can be reformulated as follows:
Cooperativity: Ensure that your speech acts in the conversation contribute opti-
mally to the salient goal of the conversation.22
Publicity: Ensure that it is common knowledge which speech acts you perform
in the conversation and what your goal for the conversation is.
These revised principles are well-defined in cases where the conversational goals
of speaker and audience conflict, which will allow these principles to serve as the
background against which we can formulate our account of manipulative speech.
gic speech. It is consistent with the idea, strongly suggested by the Stalnakerian framework, that
the goal of strategic conversation is still to share knowledge, but that the goal is not one that
need be shared by all conversational participants. The Stalnakerian can endorse this revision by
insisting that the salient goal is to share knowledge. Is also consistent with approaches that do
not treat conversation itself as having a goal, but instead treat participants as having distinct,
potentially conflicting goals. This is the approach taken, for instance, by Asher and Lascarides
[2013], who claim that participants in a conversation are being cooperative if and only if, and
to the degree to which, their goals for the conversation are aligned.
22 Defining Cooperativity as acting to further the goals of the audience is of a piece with
work in game-theoretic pragmatics, where it is common to assume that the utilities of the
audience are what fix the question under discussion [van Rooij, 2003, Franke, 2009].
14 · Justin D’Ambrosio
[S1] indicates that a strategic speaker utters some words intending to bring about
a particular action. While I think it is plausible that strategic speech is oriented
toward action, [S1] can be generalized so that it applies to any effect that a speech
act can have—any perlocutionary effect. [S2] spells out the idea that strategic
speech involves a deliberate violation of Cooperativity. If the speaker knows
that the audience’s goal for the conversation is G, the speaker utters U intending
to be less than fully cooperative with G—deliberately making U non-optimal
for pursuing G—in order to achieve the intended perlocutionary effect. Finally,
[S3] spells out the reasons the speaker is less than fully cooperative with G: it
is due to the fact that the speaker has some other goal, G ∗ , satisfaction of which
conflicts with satisfaction of G. Accordingly, the speaker intends to bring about a
perlocutionary effect A that furthers G ∗ by speaking in a way that conflicts with
G. We can then say that a speaker speaks strategically if she utters words with a
strategic linguistic intention.
In the three cases just discussed, it is clear that each of these conditions is
met. And on certain ways of spelling out the details of our examples, it may be
the case that the speakers are merely speaking strategically. However, as we will
see presently, on the most natural understanding of these cases, each speaker is
not only speaking strategically, but also trying to manipulate their audience. I
now turn to the question of what distinguishes manipulative speech from mere
strategic speech, and delay further discussion of our examples until §3.4.
A covert speech act is a speech act that can bring about its perlocutionary
aim only if it is not recognized by the audience as being that speech act.
[Saul, 2018]
A token speech act is covert if and only if it can bring about its perlocu-
tionary aim only if the goal with which that speech act is undertaken is not
recognized by the audience.
To see how these definitions differ, recall our politician from Town Hall. In Town
Hall, the politician may well hide his goal in undertaking a speech act without
hiding which speech act he undertakes. For instance, the politician might strate-
gically assert something true, but very general, in the hope of undertaking the
fewest commitments possible. This speech act token can be covert in the second
sense—he may attempt to hide that he is trying to avoid commitment—even if he
is recognized as making an assertion, and even if assertion is not a covert speech
act in Saul’s sense.
Any token of a speech act type that is covert in Saul’s sense must be covert
in my sense as well. Whenever a speech act must go unrecognized in order to
be successful, it must work toward a hidden goal. But as the case just discussed
shows, the converse is not true. Thus, the second definition of covertness sub-
sumes the first as a special case. Manipulative speech is covert in this second,
more general sense. In manipulative speech, what is hidden is the speaker’s goal
in undertaking that speech act, and whether this goal conflicts with the salient
goal of the conversation.
Before seeing how this proposal can be spelled out, one might rightly ques-
tion whether manipulative speech—and indeed, whether manipulation more
generally—is always covert. There appear to be many instances of manipulation,
in speech and otherwise, that are not covert, and a range of theorists working
on manipulation hold that covertness is not necessary for manipulation.23 Con-
23 Barnhill [2014], for instance, defines manipulation as “directly influencing someone’s
beliefs, desires, or emotions such that she falls short of ideals for belief, desire, or emotion in
ways typically not in her self-interest or likely not in her self-interest in the present context.”
16 · Justin D’Ambrosio
lying—as we will see below, the latter us a special case of the former. In cases of bald-faced
lying, it is common knowledge that the speaker is being uncooperative by asserting what they
know to be false. In cases of bald-faced manipulation, it is likewise common knowledge that
the speaker is speaking strategically, and it is common knowledge how they are doing so.
A Theory of Manipulative Speech · 17
We can then say that a speaker speaks manipulatively if and only if she utters
words with a manipulative linguistic intention. This definition is exactly the def-
inition of strategic speech, except for the addition of [M2b].25 Clause [M2b] adds
the condition that in uttering U, the speaker is attempting to get the audience
to believe that she is being fully cooperative with respect to the salient goal of
the conversation. In getting the audience to believe that her utterance is fully
cooperative, the speaker thereby hides the fact that her utterance is undertaken
to achieve some goal that conflicts with the audience’s goal for the conversation.
Clause [M2b] introduces the idea that a key feature of manipulative speech
is deceiving one’s interlocutor concerning one’s goals for the conversation. We
can see this if we return to Antique Sale. In Antique Sale, I am aware that your
goal for the conversation is to know the true value of the antique—that is the
conversation’s salient goal. In speaking manipulatively, I speak in a way that
is aimed toward getting you to pay me an exorbinant amount for an antique—
I speak strategically. But I also attempt to get you to think that my speech is
purely, or at least primarily, non-strategic: I aim to get you to think that I share
your goal of coming to have full information concerning the antique’s value, and
speak in a way that helps you achieve this goal.26 In order to accomplish these
things, I may deliberately speak vaguely, lie, mislead, or exaggerate, all while
attempting to hide that my speech is deliberately non-cooperative, and so make
myself appear ingenuous.
Turning now to Town Hall, on the view just proposed, the politician’s speech
is manipulative when it is not fully cooperative with the audience’s goal of get-
ting information concerning his plans, but he intends, and takes steps to ensure,
that the audience thinks that he is being fully revealing. This will often be impor-
tant. If, as we suggested above, it is rational for the politician to be strategically
vague, ambiguous, or general, it is very much to his advantage if he is not recog-
25 Weaker formulations of [M2b] are possible. Rather than all-out beliefs, it is plausible that
audiences merely have credences concerning speakers’ goals and the strategies they deploy.
Thus we might phrase [M2b] instead as the intention that the audience have a sufficiently high
confidence that the speaker is being fully cooperative, where what qualifies as sufficiently high
will depend on the audience or receiver’s utility function.
26 What is the salient goal of conversation when an audience who is being manipulated
contributes to a conversation, and so becomes the speaker? The answer is: it is undefined. Since
the audience is deceived about the speaker’s goal—and thinks it aligns with their own—there
is no salient goal. Likewise, Cooperativity will be undefined, since it is defined in terms of the
salient goal. But this is as it should be. One cannot be cooperative with an interlocutor if one is
deceived about their goals. One can, however, speak optimally with respect to the goals one
believes they have.
18 · Justin D’Ambrosio
nized as being so. If he is so recognized, his speech will often not achieve its aim;
there are costs to being recognized as strategic. Thus the politician will disguise
his strategies. He can do this in any number of ways: by offering assurances
that he is being non-strategic, by speaking in ways that indicate he shares the
audience’s goals, by insisting that he is a straight-shooter, and even by speak-
ing erratically or in a down-home manner. All of these tactics can disguise the
fact that his speech is intentionally non-optimal for pursuing the salient goal of
sharing knowledge concerning the question at hand.
Now consider Cross-Exam. Above we said only that in Cross-Exam, Bronston
is being strategic by not offering a direct answer to the prosecutor’s question.
But Bronston’s response is also manipulative. Why? Clearly it is advantageous
for Bronston to hide that his speech is strategic. If Bronston can convince the
prosecutor that he is being fully cooperative, then the prosecutor will take his
statement concerning the company to implicate a negative answer to the pros-
ecutor’s question, as opposed to merely being an intentionally irrelevant claim
made to misdirect.27 This is important for Bronston. Without the assumption
that Bronston is being cooperative, the prosecutor will treat his response as ir-
relevant, and respond by saying “You haven’t answered the question.” Thus, by
hiding his non-cooperativity, and trying to get the prosecutor to think that he is
furthering the salient goal, Bronston aims first to get the prosecutor to interpret
his response as implicating a negative answer, and then, if things work out for
him, to believe what he has implicated.
In each of these cases, the speaker intends the audience to believe that she
shares their goal for the conversation, and in so doing she tries to hide her ulte-
rior goal. Against this background of deception, the speaker speaks in a way that
aims to further her ulterior goal. In Cross-Exam, Bronston attempts to further
his goal by getting the prosecutor to calculate a false implicature. But not all in-
stances of a speaker being covertly uncooperative trigger a process of pragmatic
repair. Whether an audience calculates an implicature will depend on whether
the speaker is obviously or subtly non-cooperative with what the audience takes
the speaker’s goal to be. It is plausible that the speakers in Town Hall and An-
tique Sale are subtly non-cooperative; they attempt to further their ulterior goals
without an obvious violation of a maxim. Grice [1989a] himself points out this
form of manipulative speech when he says that a speaker may “quietly and un-
ostentatiously violate a maxim; if so, in some cases he will be liable to mislead”
[Grice, 1989a, p. 30].
To illustrate, suppose that in Town Hall, the politician says something am-
27 Asher
[2012], Asher and Lascarides [2013] discuss the case of implicatures in strategic
contexts. They introduce a notion called “safety” that specifies when it is reasonable for an
audience to take a speaker to have implicated something. When the audience’s assessment of
the speaker’s cooperavity is high enough, it will be safe to calculate an implicature.
A Theory of Manipulative Speech · 19
manner, his accent, his colloquial diction, and even his offensive missteps. Each
of these mannerisms contributes to the idea that Trump’s speech is nonstrategic.
And Trump isn’t the only offender—his performance is just the most obvious.
among others. One important question that carries over from the literature on trust is whether
22 · Justin D’Ambrosio
my purposes it suffices to point out that just as speaking truly and with good
reason is one of many components of cooperativity, trusting that a speaker is
speaking truly and with good reason is one of many components of linguistic
trust. Further, just as the other components of cooperativity are roughly captured
by the Gricean maxims, there are components of linguistic trust corresponding
to each of the Gricean maxims. Linguistic trust is thus a broad notion of trust
that plausibly subsumes issues of trust related to truth and justification as special
cases.
The definition above has a range of important consequences for the philosophy of
language. Perhaps the most important consequence is that it provides us with a
simple formula that makes clear predictions about which forms of speech qualify
as manipulative. The formula proceeds as follows. First, we consider all of the
forms of strategic speech. Here, since Grice’s maxims provide a rough guide to
what qualifies as cooperative speech, the maxims likewise serve as a rough guide
to the various forms that strategic speech can take. The definition then makes
the following prediction: any way of deliberately violating a Gricean maxim in
service of a hidden goal that conflicts with the salient goal of the conversation
will be an instance of manipulative speech.
5.1. Lying
This definition correctly classifies lying as a form of manipulative speech. We
can define lying as a form of speech in which a speaker deliberately asserts some-
thing concerning the topic of discussion whose content she knows to be false.31
Lying then meets our definition of manipulative speech. Why? The key reason is
that in asserting something that she knows to be false, the speaker has violated
the norm of Cooperativity. One of Grice’s maxims is the maxim of Quality, which
exhorts us to be truthful. But in standard cases of lying, a speaker intends for
their assertion to be accepted as true, and so intends for their assertion to be seen
as cooperative—they intend that their interlocutor believe that in making that as-
sertion, they are being fully cooperative with the salient goal of the conversation.
Thus, the speaker has a manipulative linguistic intention, and so lying is a form
of manipulative speech.
linguistic trust is rational. This is an important question for the Gricean: is the presumption of
Cooperativity always rational? Here I will remain neutral.
31 For discussion of lying, see Fallis [2009], Saul [2012], and Stokke [2013], among others.
A Theory of Manipulative Speech · 23
5.2. Misleading
Next, consider an oft-discussed case of misleading through implicature, similar
to the case of Bronston and the prosecutor above. Justin and Janet are dating, but
Justin, being the jealous type, is concerned that Janet is seeing her ex, Valentino,
and so asks about what she’s been doing:
(2) a. Justin: Have you been seeing Valentino this past week?
b. Janet: Valentino has mononucleosis. [Asher and Lascarides, 2013]
If Justin trusts that Janet is being cooperative, Justin will calculate the implicature
that she hasn’t been seeing Valentino. But if Janet is working toward a conflicting
goal—namely, hiding that she has been seeing Valentino—then Justin shouldn’t
calculate the implicature—his trust is unwarranted, and he should treat her re-
sponse as irrelevant. If Janet wants Justin to calculate the implicature, as opposed
to treating her response as irrelevant, it is rational for her to attempt to appear
cooperative. Thus, given Janet’s goal, it is rational for her to speak in a way
that conflicts with Justin’s goal of coming to know whether she has been seeing
Valentino, while also hiding that her speech is not fully cooperative, in order to
get Justin to calculate and accept a false implicature. In doing so, Janet meets the
conditions laid out in our definition of manipulative speech above. Misleading
is a form of manipulation.
5.3. Dogwhistling
In her important discussion of dogwhistles, Saul [2018] classifies dogwhistles as
a key form of political manipulation. The definition above yields the same result,
at least concerning what Saul calls covert dogwhistles. To see this, consider the
following example:
(3) “Yet there’s power, wonder-working power, in the goodness and idealism
and faith of the American people.” –George W. Bush
Here, Bush uses a covert dogwhistle to signal to his Evangelical constituency that
he is intending to endorse fundamentalist Christian values. But his use of this
dogwhistle is clearly strategic, and clearly covert. There are many different com-
peting accounts of dogwhistles, but for our purposes we can treat dogwhistles
as utterances that admit of two or more interpretations, one of which is intended
to convey something distinctive to a select subset of an audience.32
32 Thus, I here remain neutral on whether the two available contents are semantically encoded
or merely conveyed. Stanley [2015] claims that dogwhistles involve not-at-issue content that
is semantically encoded. By contrast, Khoo [2017] treats code words as conveying different
contents through a pragmatic mechanism. Henderson and McCready [2018] do likewise.
24 · Justin D’Ambrosio
avoid the costs associated with being detected as strategic, and so make it more
likely that his are accepted by the audience. This form of manipulative speech
that is extremely common in political speech.
But there are many examples of manipulative underspecification outside of
politics. Horoscope writers intend their audiences to believe that they are making
precise, substantive predictions, when in fact they are using only vague language
that rules almost nothing out. The same is true for fortune-tellers, who aim to
get their clients to believe that they have made specific predictions—and so have
been fully cooperative—when in fact they have covertly been as vague as possi-
ble, couching their predictions in terminology that suggests specificity. The same
is often true for many forms of prediction: in sports, economics, the weather, and
even medicine.
6. Propaganda
Turning now to political speech, the theory developed above can be used to for-
mulate a novel, attractive account of the nature of propaganda and its mecha-
nisms. First, note that it is standard to define propaganda in terms of manipula-
tion. Consider, for example, the following definitions:
While these definitions differ in important ways, each of them claims, in as many
words, that propaganda operates by means of manipulation, and plausibly, the
primary medium through which propaganda operates is through speech. Thus,
manipulative speech is at least a core component of propaganda. But what makes
propaganda distinctive?
A Theory of Manipulative Speech · 27
One key way in which propaganda differs from other forms of manipulative
speech is that propaganda does not seem to occur as part of a conversation. This
is due to the fact, also apparent in the definitions above, that propaganda is
issued from a single source, but aimed at a mass audience. This points us to
an initial proposal about the connection between propaganda and manipulative
speech:
This proposal, while appealing, is too narrow. While speech is perhaps the pri-
mary medium through which propaganda operates, it is surely not the only
one. There can be propagandistic signs, music, art, film, and architecture—
propaganda can operate through a wide range of media. Moreover, it would
be useful to know what changes, if any, must be made to our account of manip-
ulative speech once such speech is directed at a mass audience.
In order to formluate a better proposal, we can here follow Stanley [2015] in
introducing the idea of public discourse. Public discourse can be thought of as a
generalization of conversation to the case where a speaker speaks to a mass au-
dience. Public discourse retains many, but not all of conversation’s key features.
Contributors to public discourse will likewise have goals, and such goals—along
with the ways they are pursued—can be hidden. Members of audiences, likewise,
will have goals, and those may align more or less fully with those of the speaker.
Each member of the audience will likewise have an assessment of how fully a
speaker’s goals align with their own: they will have some degree of linguistic
trust.
But there are key differences between ordinary conversation and public dis-
course. First, public discourse, unlike conversation, is often one-sided. Audi-
ences in public discourse often do not have the chance to respond to speakers.
Further, in public discourse, the goals and background beliefs of mass audiences
are not uniform, and members of the audience will not all have the same degree
of linguistic trust in the speaker. This variation will lead to different strategic
choices on the part of the manipulative contributors to public discourse, and dif-
ferent ways of cultivating linguistic trust. Speakers will have to operate with a
summary assessment of the goals of their audience, or perhaps with multiple
assessments of the goals of different parts of the audience. Finally, unlike in
conversation, contributions to public discourse need not be linguistic. One can
contribute to public discourse by making a movie, composing a piece of music,
hanging a banner, or in any number of other ways.
With the idea of a contribution to public discourse in hand, we can offer the
following definition of propaganda:
When combined with the theory of manipulative speech developed above, this
proposal significantly advances our understanding of propaganda. The account
developed above provides us with a precise characterization of what it is for a
contribution to conversation to be manipulative: a contribution to a conversa-
tion is manipulative if it is covertly strategic. But this definition can be straight-
forwardly extended to contributions to public discourse. Just as in the case of
manipulative speech in conversation, propagandists speak so as to pursue goals
that they know conflict with the goals of the public, all while trying to make their
speech appear cooperative with the public’s goals.
Of course, in light of the differences between conversation and public dis-
course, the speaker will make use of a more complicated assessment of the au-
dience in choosing her manipulative strategies. But the basic structure of ma-
nipulative speech and propaganda is the same. Propagandists will do whatever
they can to convince public audiences that they share the public’s goals, and so
that their contributions to public discourse are intended to further those goals.
The propagandist will then be successful to the degree that the audience inter-
prets those contributions in a way that brings about the propagandist’s intended
perlocutionary effect.
Like manipulative speech in conversation, propaganda operates by exploit-
ing the public’s trust that a contribution is made against a background of aligned
goals. Accordingly, many of the same mechanisms of cultivating and exploiting
linguistic trust that were effective in the case of conversation will be effective
here. Propagandists can pose as non-strategic by speaking in ways that make
them appear ingenuous, by appearing to endorse the goals that they know the
audience holds, or by speaking in ways that foster linguistic trust. Thus, a key
element of propaganda is deceiving a public audience about one’s goals. Pro-
paganda is a covertly non-cooperative contribution to public discourse aimed at
bringing about a particular—often political—perlocutionary effect.
This view of propaganda subsumes a wide range of instances of propaganda
discussed in the literature, and improves on prior definitions. In a recent discus-
sion, Stanley [2015] distinguishes between several different forms of propaganda:
supporting propaganda, undermining propaganda, and demagoguery. Consider
first his definition of undermining propaganda:
other. If the audience does not share the goals or ideals that the contribution
to public discourse appears to advance, then the propaganda cannot hope bring
about its perlocutionary aims.
Stanley appears to recognize this, and goes on to define a form of propaganda
that he calls “undermining demagoguery”:
7. Conclusion
The above theory provides us with a framework in which to study the varieties
of manipulative speech. But it also leaves much work still to be done. First,
there are important questions concerning how to model covertly strategic speech.
Frameworks that model contexts as bodies of common knowledge, or as shared
conversational scoreboards, are not adequate to the task, because in manipulative
30 · Justin D’Ambrosio
speech, speaker and audience have different conceptions of how the conversa-
tion is going. To account for this, we need scoreboards that can model deception
concerning a speaker’s goals and other facts about the context relevant to lin-
guistic interpretation. Further, while I have here discussed a few speech acts
predicted to be manipulative by this definition, there are many forms of manip-
ulative speech still to be studied.
Perhaps more importantly, the theory provides us with the beginnings of
a toolkit for detecting—and so resisting—manipulative speech in politics and
public life. Manipulative speakers exploit our presumptions of cooperativity
to deceive us about their goals, and in so doing make it more likely that we
accept their often nefarious claims. By investigating the nature of linguistic trust,
examining our default presumptions of cooperativity, and paying close attention
to the goals of public figures and the language they use to pursue them, we can
cast light on strategic speech intended to be covert. In so doing, we will see that
many figures in public life do not speak in an honest, straightforward manner.
Rather, like Marc Antony, they do not speak right on, and stir our blood only by
convincing us that they do.
References
Don Fallis and Andreas Stokke. Bullshitting, lying, and indifference toward truth.
Ergo, 4, 2017.
Joseph Farrell and Matthew Rabin. Cheap talk. Journal of Economic Perspectives,
10(3):103–118, Summer 1996.
Yossi Feinberg. Strategic communication. Proceedings of TARK XIII, pages 1–11,
July 2011.
Michael Franke. From Signal to Act: Game Theory in Pragmatics. PhD thesis, Insti-
tute for Logic, Language, and Computation, University of Amsterdam, Ams-
terdam, the Netherlands, 2009.
Michael Franke and Robert van Rooij. Strategies of persuasion, manipulation,
and propaganda: Psychological and social aspects. In Sujata Gosh Johan van
Benthem and Rineke Verbrugge, editors, Models of Strategic Reasoning: Logics,
Games, and Communities, chapter 8. Springer, Heidelberg, 2015.
Harry G. Frankfurt. On Bullshit. Princeton University Press, Princeton, NJ, 2005.
Elizabeth Fricker. Stating and insinuating. The Aristotelian Society Supplementary
Volume, 86(1):61–94, 2012.
Katherine Frost-Arnold. The cognitive attitude of rational trust. Synthese, 191(9),
2014.
Katherine Frost-Arnold. Social media, trust, and the epistemology of prejudice.
Social Epistemology, 30(5-6):513–531, 2016.
Sanford Goldberg. Self-trust and extended trust: A reliabilist account. Res Philo-
sophica, 90(2):277–292, 2013.
Sanford C. Goldberg. Relying on Others: An Essay in Epistemology. Oxford Uni-
versity Press, Oxford, 2012.
Robert E. Goodin. Manipulatory Politics. Yale University Press, Yale and London,
1982.
Paul Grice. Studies in the Way of Words. Harvard University Press, 1989a.
Paul Grice. Logic and conversation. In Studies in the Way of Words Grice [1989a],
pages 22–40.
Jason Hanna. Libertarian paternalism, manipulation, and the shaping of prefer-
ences. Social Theory and Practice, 41(4):618–643, 2015.
Katherine Hawley. Trust, distrust and commitment. Noûs, 48(1):1–20, 2014.
Katherine Hawley. How to be Trustworthy. Oxford University Press, Oxford, 2019.
Robert Henderson and Elin McCready. How dogwhistles work. Proceedings of
LENLS, 2018.
Robert Henderson and Elin McCready. Dog whistles and the at-issue/non-at-
issue distinction. In Daniel Gutzman and Katharina Turgay, editors, Secondary
Content: The Semantics and Pragmatics of Side-Issues, volume 37 of Current Re-
search in the Semantics/Pragmatics Interface, pages 222–245. Brill, Leiden, The
Netherlands, 2019.
Edward S. Herman and Noam Chomsky. Manufacturing Consent. Pantheon
A Theory of Manipulative Speech · 33
Anthony Pratkanis and Elliot Aronson. Age of Propaganda: The Everyday Use and
Abuse of Persuasion. W. H. Freeman and Co, 1992.
Anne Quaranto and Jason Stanley. Propaganda. In Justin Khoo and Rachel
Sterken, editors, Routledge Handbook of Social and Political Philosophy of Lan-
guage. Routledge, forthcoming.
Joseph Raz. The Morality of Freedom. Oxford University Press, Oxford and New
York, 1988.
Craige Roberts. Information structure in discourse: Towards an integrated for-
mal theory of pragmatics. In Jae-Hak Yoon & Andreas Kathol, editor, OSU
Working Papers in Linguistics, Vol 49: Papers in Semantics. The Ohio State Uni-
versity Department of Linguistics, 1996.
Jennifer Saul. Dog whistles, political manipulation, and the philosophy of lan-
guage. In Daniel W. Harris Daniel Fogal and Matt Moss, editors, New Work
on Speech Acts. Oxford University Press, 2018.
Jennifer M. Saul. Lying, Misleading, and What is Said: An Exploration in Philosophy
of Language and Ethics. Oxford University Press, Oxford and New York, 2012.
T.M. Scanlon. What We Owe to Each Other. Harvard University Press, Cambridge,
MA, 1998.
Lawrence M. Solan and Peter M. Tiersma. Speaking of Crime: The Language of
Criminal Justice. Chicago Series in Law and Society. University of Chicago
Press, Chicago, 2005.
Robert Stalnaker. Context. Oxford University Press, Oxford and New York, 2014.
Robert C. Stalnaker. On the representation of context. Journal of Logic, Language,
and Information, 7:3–19, 1998.
Robert C. Stalnaker. Common ground. Linguistics and Philosophy, 25:701–721,
2002.
Jason Stanley. How Propaganda Works. Princeton University Press, Princeton, NJ,
2015.
Andreas Stokke. Lying, deceiving, and misleading. Philosophy Compass, 8(4):
348–359, 2013.
Cass Sunstein. The Ethics of Influence: Understanding and Dealing With Manip-
ulative People. Parkhurst Brothers Publishers Inc., Little Rock, AR, revised
edition edition, 2016.
Zoltán Gendler Szabó. The goal of conversation. Aristotelian Society Supplemen-
tary Volume, 94(1):57–86, 2020. Manuscript, Joint Meeting of the Mind and
Aristotelian Society.
Michael Tomz and Robert P. van Houweling. The electoral implications of can-
didate ambiguity. American Political Science Review, 103(1):83–98, February
2009.
Robert van Rooij. Questioning to resolve decision problems. Linguistics and Phi-
losophy, 29:727–763, 2003.