§1

Assertion and its Constitutive Norms

Michael Rescorla

Abstract: Alston, Searle, and Williamson advocate the restrictive model of assertion, according

to which certain constitutive assertoric norms restrict which propositions one may assert. Sellars

and Brandom advocate the dialectical model of assertion, which treats assertion as constituted by

its role in the game of giving and asking for reasons. Sellars and Brandom develop a restrictive

version of the dialectical model. I explore a non-restrictive version of the dialectical model. On

such a view, constitutive assertoric norms constrain how one must react if an interlocutor

challenges one’s assertion, but they do not constrain what one should assert in the first place. I

argue that the non-restrictive dialectical perspective can accommodate various linguistic

phenomena commonly taken to support the restrictive model.

§1. Constitutive assertoric norms

Assertion invites normative assessment from diverse perspectives. We may condemn

Iago’s lies to Othello as immoral, simultaneously commending how effectively they promote

Iago’s goals or how courteously Iago expresses himself. In rendering these judgments, we deploy

evaluative standards rooted in morality, instrumental rationality, and propriety. The evaluative

standards exhibit no special connection to language. We could readily apply them to Iago’s non-

linguistic actions. Assertion falls under general norms lacking any peculiarly linguistic character.

Does assertion also fall under peculiarly linguistic norms? Norms that issue, not from

general principles of morality, rationality, or social interaction, but from special features of

linguistic intercourse? Many philosophers answer affirmatively, arguing that assertion falls under

constitutive norms. On this view, assertion is analogous to a game, in that it intrinsically involves

certain rules. Someone who exhibits no sensitivity to the rules of a game does not grasp what it is

to play the game, and someone who exhibits no sensitivity to assertion’s constitutive norms does

not grasp what it is to assert a proposition. Proponents of this view include Alston (2000),

Brandom (1994), Douven (2006), Searle (1969, 2001), Sellars (1963), and Williamson (2000).

Williamson holds that assertion falls under a constitutive norm with the structure: One

should assert p only if C(p). Williamson’s candidate for ‘C(p)’ is ‘one knows p,’ yielding

The Knowledge Norm: One should assert only propositions that one knows.

Countless other norms with the same structure are possible, including:

The Honesty Norm: One should assert only propositions that one believes.

The Truth Norm: One should assert only true propositions.

The Warrant Norm: One should assert only propositions that one believes

with ‘sufficient’ warrant.

I will call such norms restrictive, since they restrict which propositions one may assert. The

restrictive model of assertion holds that assertion falls under constitutive restrictive norms.

Alston, Brandom, Searle, and Sellars endorse the restrictive model, as do most philosophers who

embrace constitutive assertoric norms.

One can also imagine non-restrictive assertoric norms. I focus on norms with the

structure: If one asserts p, then, if D(p), one must φ(p). Theories based entirely upon norms with

this structure reject any notion of ‘assertibility’ or ‘appropriate assertion.’ They characterize

assertion solely in terms of how it alters the speaker’s normative standing, not in terms of which

assertions are permissible.

The most promising candidates for non-restrictive assertoric norms reflect what I will call

the dialectical model of assertion, which regards assertion as essentially a move within ‘the

game of giving and asking for reasons.’ The intuitive idea here is that, by asserting a proposition,

I commit myself to defending the proposition when faced with challenges and counter-

arguments. I can cancel the dialectical commitment by retracting my assertion, but until then the

commitment stands. Although the basic idea is ancient, recent discussion of it derives mainly

from the writings of Hamblin (1971), Sellars (1963), and Toulmin (1958). However we elaborate

this idea, the resulting account will likely feature non-restrictive norms that describe how to react

if another speaker contests an asserted proposition.

One can develop the dialectical model in either a restrictive or non-restrictive direction.

On the restrictive version, constitutive assertoric norms codify how to answer challenges and

counter-arguments, and they also constrain which propositions one can initially assert. Brandom

(1994), Sellars (1963), and Watson (2004) advocate views along these lines. I will explore a non-

restrictive version of the dialectical model. On this view, constitutive assertoric norms constrain

how I must react if someone challenges my assertion, but they do not constrain what I should

assert in the first place. Morality, propriety, or rationality might require that I assert only

propositions with some desired feature. No constitutive norm of assertion imposes any such

requirement. For instance, a lie violates no constitutive assertoric norm, although the liar incurs

dialectical commitments that she may or may not be able to discharge.

A non-restrictive version of the dialectical model may appear somewhat implausible. I

am aware of only a single recent commentator, MacFarlane (2003, 2005), who advocates

anything resembling it. I will urge that some popular arguments for the restrictive model are

seriously flawed. I will furthermore argue that the non-restrictive dialectical perspective can

accommodate various linguistic phenomena commonly taken to support the restrictive model.

§2. Norms and practices

In this section, I develop a rough conception of how constitutive assertoric norms differ

from non-constitutive assertoric norms. I hope that the conception will be fairly uncontroversial.

Advocates of constitutive assertoric norms typically emphasize the analogy with games,

as in the following passage by Williamson: “one might suppose… that someone who knowingly

asserts a falsehood has thereby broken a rule of assertion, much as if he had broken a rule of a

game; he has cheated. On this view, the speech act, like a game and unlike the act of jumping, is

constituted by rules” (p. 238). The analogy with games seems helpful, so we should explore it in

detail. How exactly is assertion more analogous to a game than to an activity such as jumping?

The contrast between playing a game and jumping illustrates a more general contrast

between what I will call practices and mere activities. Besides games, sample practices include:

dances, like the waltz or the tango; religious ceremonies, like weddings; fraternity initiation rites;

performing Beethoven’s Fifth Symphony. Every practice is associated with “internal” standards

of normative assessment codified by norms dictating how to execute the practice correctly. The

norms for performing Beethoven’s Fifth Symphony demand that musicians play the notes of

Beethoven’s score at roughly the indicated tempo. The norms for dancing the waltz demand that

dancers move in time with the music, selecting their movements from a fixed repertoire of dance

steps. Agents who violate a practice’s norms do not implement the practice correctly, but they

may still implement the practice. A devious gambler might repeatedly cheat during a poker

game. He plays poker incorrectly, but he plays poker nonetheless.

The connection with internal evaluative standards differentiates practices from mere

activities, like jumping, bathing, or holding hands. A mere activity falls under general norms,

like those of morality, propriety, and rationality, but it does not fall under specialized norms that

govern correct execution of the activity. There is no correct or incorrect way to bathe. There is

no correct or incorrect way to hold hands.

The contrast between practices and mere activities suggests the following formulation: a

norm is constitutive of a practice iff one must obey the norm to engage correctly in the practice.

In some weak sense, any activity involves constitutive requirements: requirements one

must satisfy to count as engaging in the activity. One must immerse parts of one’s body in some

liquid to count as ‘bathing.’ Two people must achieve physical contact between their hands to

count as ‘holding hands.’ But only practices enshrine constitutive standards that govern correct

execution of the activity. Any practice engenders a three-fold division between actions that do

not count as engaging in the practice, actions that count as engaging in it correctly, and actions

that count as engaging in it incorrectly. A mere activity engenders only a two-fold division

between actions that count as engaging in the activity and actions that do not. We might describe

this contrast by saying that all activities involve constitutive requirements, which one must

satisfy to engage in the activity, while practices also involve constitutive norms, which one must

satisfy to engage in the practice correctly.

There are basically two ways one might analyze constitutive assertoric norms within the

framework developed thus far. Most straightforwardly, we might treat assertion as a self-

contained practice governed by constitutive norms. On this view, constitutive assertoric norms

describe how to engage correctly in the practice of assertion, where that practice can be

characterized in isolation from other speech acts. Alston, Searle, and Williamson adopt this

approach. A somewhat less straightforward approach treats assertion as a move within a larger

practice, much as a home run is a move within the larger practice of baseball. On this view,

constitutive assertoric norms describe how to engage correctly in some segment of the larger

practice. Sellars and Brandom adopt this approach, insisting that assertion is individuated by its

role within the game of giving and asking for reasons, which I will often refer to as ‘reasoned

discourse.’ If Sellars and Brandom are correct, then assertion can be fully characterized only

through its normative relations to other speech acts, such as questioning and challenging.

Although I will side with Sellars and Brandom in this dispute, our terminology should not

take sides. I offer the following characterization. A norm is a norm of assertion iff my assertoric

performances are relevant to whether I violate it. A norm is constitutive of assertion iff it is a

norm of assertion and it is constitutive of a practice to which assertion is intrinsically connected.

This formulation is rough, but it is general enough to encompass Alston, Searle, and Williamson

(since the practice might simply be assertion itself) and Sellars and Brandom (since the practice

might be the game of giving and asking for reasons). Our formulation reflects the intuitive idea

that assertion has an intrinsically normative dimension. We can characterize assertion’s

fundamental nature only by delineating specialized norms that govern its correct execution. I

think this conception guides many philosophers, either more or less explicitly.

On the other hand, many philosophers repudiate specialized assertoric norms. Davidson

is a particularly emphatic opponent (1984, pp. 265-280). In my terminology, Davidson regards

conversation as a mere activity, rather than as a practice incorporating its own internal standards

of normative assessment. For Davidson, there is no significant distinction between ‘correctly’ or

‘incorrectly’ executing speech acts. One either executes them or one does not. Such an approach

precludes anything like constitutive norms of speech acts. It countenances only whatever

normativity results from subsuming conversation under extra-linguistic norms. On Davidson’s

approach, assertion is not intrinsically normative, although, like any action, it falls under norms

of morality, propriety, rationality, etc.

My discussion highlights the elementary but crucial distinction between a norm’s content

and its normative status. When discussing ‘norms of assertion,’ it is never enough merely to

specify a norm’s content, i.e. what the norm demands. One must also specify what normative

force it exerts. In some sense, it is clearly a ‘norm of assertion’ that one ‘should’ speak honestly.

The question is whether the Honesty Norm enjoys a specifically linguistic grounding beyond

what it receives from morality, legality, propriety, rationality, and other extra-linguistic factors.

Is the Honesty Norm constitutive of assertion? Or does it merely result from subsuming assertion

under more general normative patterns?

§3. The dialectical model of assertion

According to the dialectical model, assertion involves a commitment to defend the

asserted proposition if challenged. The model develops this idea, in part, by positing constitutive

norms that regulate how one should respond when other speakers challenge one’s assertions.

Does ‘defending’ my assertion require arguments that actually support my position, or

that I regard as doing so, that my interlocutor regards as doing so, or something else? Can my

argument employ any non-question-begging premises, or only true premises, or only premises

my interlocutor accepts, or something else? How we answer such questions determines which

norms we embrace. For instance, selecting the first answer to each question suggests

The Defense Norm: When challenged to defend an asserted proposition, one must

either provide a cogent, non-circular argument for the proposition or else retract it,

where a ‘cogent’ argument is one whose premises provide rational support for its conclusion.

Often, proponents of the dialectical model regard the Defense Norm as too strong, preferring

instead something closer to

The Default-Challenge Norm: When faced with a legitimate challenge to defend an

asserted proposition, one must either provide a cogent, non-circular argument for the

proposition or else retract it.

For example, a challenge to the assertion “I have hands” might count as legitimate only if the

challenger sketches a compelling skeptical scenario in which I do not have hands. Leite (2005)

presents a theory of reasoned discourse based on something like the Default-Challenge Norm. He

does not endorse the further claim that this norm is constitutive of assertion. Brandom endorses

the further claim. In (Rescorla, 2009b), I defend a weakened variant of the Defense Norm against

Leite and Brandom. We may ignore these internecine skirmishes here.

The dialectical model also describes how one should respond to counter-arguments

against an asserted proposition. Again, many variations are possible. Most accounts feature

something like the following norm:

The Retraction Norm: When faced with a counter-argument against an asserted

proposition, one must rebut the counter-argument or else retract the proposition.

Note that we can accept the Retraction Norm even if we reject anything resembling the Defense

or Default-Challenge Norm, a combination of views suggested by (MacFarlane, 2005, p. 334).

If I seek to discharge a dialectical commitment, as opposed to canceling it, then I must

provide an argument. How do I provide an argument? By asserting additional propositions as the

argument’s premises. Thus, assertion plays a dual role in reasoned discourse: through it, one both

undertakes and discharges dialectical commitments. As Brandom puts it:

[A]ssertings… are in the fundamental case what reasons are asked for, and what giving a

reason always consists in. The kind of commitment that a claim of the assertional sort is

an expression of is something that can stand in need of (and so be liable to the demand

for) a reason; and it is something that can be offered as a reason (1994, p. 167).

Assertion’s dual role might appear to generate a vicious regress, since discharging dialectical

commitments embroils one in further commitments. Proponents of the dialectical model evince

various reactions to this regress. Brandom tries to halt the regress through the Default-Challenge

Norm, which allows us to say that certain assertions require defense only given special effort by

one’s interlocutor. In contrast, (Rescorla, 2009a) argues that the regress is not worrisome: either

speaker and interlocutor agree upon mutually acceptable relevant premises, in which case the

regress halts, or speaker and interlocutor do not agree upon mutually acceptable premises, in

which case the speaker leaves certain dialectical commitments undischarged.

Again, we may

ignore these complications here.

To undercut the dialectical model, some critics emphasize that serious dialectical

interaction is a relatively rare phenomenon. We often assert propositions without engaging one

another in reasoned discourse. Yet this undeniable fact is consistent with the dialectical model,

which holds that asserting a proposition is performing an action that occupies a certain normative

role within reasoned discourse. Such a view can allow that assertion sometimes occurs outside

reasoned discourse. Non-dialectical assertoric performances may even statistically outnumber

performances within reasoned discourse. The model claims only that non-dialectical assertoric

performances are explanatorily derivative from core performances within reasoned discourse.

Part of what makes a given linguistic performance an assertion, as opposed to some other speech

act, are the dialectical commitments the speaker assumes, commitments that become operative

once the speaker engages fellow conversationalists in reasoned discourse. In this sense, assertion

is ‘intrinsically connected’ to reasoned discourse.

Not all assertions are moves within reasoned discourse, but all assertions are potential

moves within reasoned discourse. Whenever I assert some proposition, other speakers may

challenge me to defend it or even provide counter-arguments against it. And then I face a choice.

I can extrude myself from reasoned discourse (“I’m not going to argue with you”). I can engage

in reasoned discourse incorrectly, trying to defend my assertion but doing so inadequately. Or I

can engage in reasoned discourse correctly, vindicating my assertion or else retracting it. Thus,

the norms of reasoned discourse describe a normative infrastructure that lurks beneath the

surface of quotidian conversation, perpetually threatening to emerge. According to the dialectical

model, assertion is individuated by its place within this normative infrastructure and hence is

possible only against the background provided by it.

But aren’t there are many assertoric performances through which I undertake no

dialectical commitment whatsoever? Suppose I engage someone in casual small-talk at a cocktail

party. I might assert a proposition to initiate a new topic or promote a diverting conversation.

Yet, one might urge, it hardly seems appropriate to greet such an assertion with challenges or

counter-arguments. So the dialectical model must be fundamentally mistaken in tying assertoric

force so closely to dialectical commitment.

In response, note that there are many elliptical gambits through which I can introduce a

new topic or promote an entertaining conversation. Instead of asserting p, I might say, “I’m

curious what you think about p.” I thereby put forward a proposition p without undertaking any

dialectical commitment. If my interlocutor responds by arguing against p, I might simply reply,

“Very interesting points.” It would be outrageous for my interlocutor to then demand, “Well,

aren’t you going to respond to my arguments? Or, at least, retract what you said?”. That demand

seems far more legitimate if I initially asserted p. Although the demand may be obnoxious or

uncouth, there is a clear sense in which it is invited by an assertion that p but not by more

elliptical ways of putting forward p. This contrast suggests the following analysis: assertion

always involves some element of dialectical commitment, but general norms of social propriety

may render it more or less appropriate to press the speaker to discharge those commitments.

Most proponents of the dialectical model combine it with the restrictive model. Brandom

treats the Warrant and Default-Challenge Norms as constitutive of assertion. Watson (2004)

develops a similar account, supplemented with a restrictive norm along these lines:

The Defensibility Norm: One should assert only propositions one can defend with

cogent, non-circular arguments.

We can imagine numerous variants, such as:

The Known Defensibility Norm: One should not assert propositions one knows

one cannot defend with cogent, non-circular arguments.

In what follows, I will investigate the prospects for a non-restrictive version of the dialectical

model. I will examine various popular arguments for the restrictive model, urging that any

uncontroversial phenomena adduced by them can be accommodated just as well by the non-

restrictive dialectical perspective. For additional development of the non-restrictive dialectical

perspective, see (Rescorla, 2007).

§4. Does the dialectical model entail the restrictive model?

One might suspect that the dialectical model collapses into the restrictive model, and

hence that the non-restrictive dialectical perspective is incoherent. If assertion involves a

commitment to defend what one says, then shouldn’t one avoid asserting propositions one cannot

defend, or at least propositions one knows one cannot defend? More explicitly, one might reason:

(A) (i) By asserting a proposition, one commits oneself to defending it.

(ii) One should not undertake commitments one knows one cannot discharge.

(iii) Hence, one should not assert propositions one knows one cannot defend.

Argument (A) purports to derive a restrictive norm from the basic intuition (i) underlying the

dialectical model. If ‘defend’ means ‘defend with cogent, non-circular arguments,’ then the

derived norm is the Known Defensibility Norm.

In evaluating (A), a central consideration is the normative force of the ‘should’ in (ii) and

(iii). Undoubtedly, one should not undertake commitments one knows one cannot keep, if one

wants to keep one’s commitments. But this establishes only that

If one wants to keep one’s commitments, one should not assert propositions one knows

one cannot defend,

which is a hypothetical norm about how to achieve a certain goal (keeping one’s commitments),

not a constitutive norm of assertion. One might instead construe (ii) as a moral norm. Under this

construal, (A) is valid only if ‘should’ in (iii) likewise expresses moral obligation, in which case

(iii) is moral norm rather than a norm describing how to execute assertion correctly.

Another idea would be to treat (ii) as somehow constitutive of commitment in general, in

which case it seems more plausible that (i) and (ii) jointly yield a constitutive norm of assertion.

But, as I will now argue, this suggestion also faces problems.

Consider Rawls on promising. Rawls holds that promising falls under a constitutive norm

that “reads roughly as follows: if one says the words ‘I promise to do X’ in the appropriate

circumstances, one is to do X, unless certain excusing conditions obtain” (1971, p. 345). Rawls

explains our obligation to keep promises as resulting from interaction between this norm and the

Principle of Fairness, a general moral prohibition against free-riding on just practices. He does

not mention a constitutive norm against promising to φ when one has no intention of φ-ing or

even when one knows one cannot φ. A natural interpretation, then, is that Rawls does not think

insincere promises violate any constitutive norm of promising. Violation occurs when one fails

to keep the promise, but not necessarily before. By promising to φ when I have no intention of φ-

ing, I may be morally culpable. But moral transgressions are not necessarily violations of

constitutive promissory norms. For instance, suppose I promise to φ while feeling quite certain I

will not be able to φ; shortly after the promise, it becomes both possible and advantageous for

me to φ, so I φ. On the proposed view, I need violate no constitutive norm of promising,

although I undoubtedly violate a moral norm.

Alston (2000) and Searle (1969) argue that promising falls under a restrictive constitutive

norm that one promise to φ only if one intends to φ. Other philosophers, such as Thomson (1990,

p. 303), question whether promising involves constitutive norms at all. I take no stand regarding

which account is correct. My point is just that there is nothing incoherent about the conception I

have attributed to Rawls. One can imagine a practice along these lines, even if that practice is not

ours, and even if there are good reasons for preferring our own. The proposed practice provides a

coherent model of commitment as regulated by norms requiring one to fulfill one’s commitments

but no norms restricting which commitments one undertakes. It thereby casts doubt upon the

thesis that (ii) is constitutive of commitment in general, and hence upon (A)’s attempt to derive a

constitutive restrictive norm from the dialectical model.

There is something defective about performing action  within practice P if I know doing

so will eventually force me to engage in P incorrectly. But that does not entail that  itself,

rather than some action subsequent to , violates P’s constitutive norms. The following principle

is plausible: engaging in a practice constitutively requires at least pretending that one seeks to

engage in it correctly. Thus, depending on the details of the case, my -ing is defective in one of

two ways. Either I do not really seek to engage in P correctly, in which case my behavior is

disingenuous, since I misleadingly pretend that I seek to engage in P correctly. Or else I seek to

engage in P correctly, in which case it is pragmatically incoherent to , since I know doing so

will eventually force me to violate P’s constitutive norms. In neither case need we posit an

additional constitutive norm, a norm violated by , beyond those constitutive norms  will

eventually induce me to violate.

The non-restrictive dialectical perspective may not yet seem plausible. But it does seem

coherent. In particular, constitutive features of commitment do not induce a collapse of the

dialectical model into the restrictive model.

§5. Representing oneself as believing/knowing that p

I can speak insincerely. It may even be common knowledge between myself and my

interlocutor that I am speaking insincerely (Rumfitt, 1995). But I cannot assert a proposition

while abandoning all pretensions towards speaking sincerely. If I say “Frank is an excellent

student” while blatantly smirking and winking, then I do not assert that Frank is an excellent

student, except perhaps in some attenuated sense that does not interest us here. My body

language subverts the display of candor that is a necessary concomitant to genuinely, non-

ironically asserting a proposition. Assertion essentially involves at least a pretense of sincerity.

It is tempting to suppose that the essential link between assertion and apparent sincerity

supports the restrictive model. If assertion essentially involves a display of sincerity, then surely

one should assert a proposition only when the display is genuine.

This reasoning is fallacious. It slides without any argument from uncontroversial claims

about assertion’s constitutive requirements to highly controversial claims about assertion’s

constitutive norms. A constitutive requirement of assertion is that one adopt a show of

conviction, just as a constitutive requirement of bathing is that one immerse part of one’s body in

liquid. In neither case does it follow that the activity involves constitutive norms: norms

specifying how to execute the activity correctly. Someone who abandons the pretense of

sincerity does not engage incorrectly in assertion. She does not assert a proposition at all. Of

course, given that assertion involves a display of sincerity, and given the moral principle that one

should avoid deceiving others, it follows that one should speak sincerely. But that still does not

show that any norm is constitutive of assertion. It only illustrates that assertion falls under

general norms that apply to many other actions.

Few philosophers would embrace the fallacious reasoning in the simplistic form just

criticized. But similar mistakes plague much more sophisticated expositions.

Unger (1975, p. 253) advocates the following thesis: “If someone asserts, states, or

declares that something is so, then it follows that he represents himself as knowing that it is so.”

Let us call this Unger’s Thesis. According to DeRose (2002, p. 180), Unger’s Thesis entails that

the Knowledge Norm is constitutive of assertion: “For our purposes, these are just two sides of

the same coin: If one represents oneself as knowing that p by asserting p, then, to avoid falsely

representing oneself, one should follow the rule of asserting only what one knows.”

I am not convinced that Unger’s Thesis is true. But the key point here is that Unger’s

Thesis does not entail that the Knowledge Norm is constitutive of assertion. At best, Unger’s

Thesis identifies a constitutive requirement of assertion: to assert p, one must represent oneself

as knowing p. That assertion embodies this constitutive requirement does not entail that there are

correct or incorrect ways of asserting a proposition. Only a confusion between constitutive

requirements and constitutive norms, or between constitutive and non-constitutive norms, could

make the inference from Unger’s thesis to the restrictive model seem obvious and unproblematic.

Davidson, who endorses a weakened version of Unger’s Thesis that replaces knowledge with

belief, puts the point succinctly: “What we do in making an assertion is represent ourselves as

believing what we say, and so we may be morally at fault if it turns out that we don’t believe

what we asserted. But moral error is not linguistic error” (2005, p. 123).

Hence, there is no deductive entailment from Unger’s Thesis, or similar doctrines, to the

restrictive model of assertion. But one might urge that some kind of abductive entailment holds.

In the previous section, we mentioned the following extremely plausible principle: engaging in a

practice constitutively requires at least pretending that one seeks to engage in it correctly. This

principle, combined with the Honesty Norm, entails that assertion constitutively requires at least

a pretense of sincerity. Similar derivations are possible for Unger’s Thesis, or for the weakened

version of it espoused by Davidson (Williamson, 2000, p. 252). So the restrictive model can

explain assertion’s constitutive requirements.

One problem with the proposed abductive argument is that one might simply deny that

assertion’s constitutive requirements need explaining. Davidson urges that his weakened version

of Unger’s Thesis follows directly from the conceptual analysis of assertion (1984, p. 270). It is

not clear that the proposed abductive argument exerts any force against Davidson’s position.

However, the abductive argument poses a challenge to the non-restrictive dialectical

perspective. The argument shows that the restrictive model easily accommodates relatively

uncontroversial features of assertion. It is not clear that a non-restrictive dialectical account can

explain those same features. Apparently, I can undertake a commitment to defend some

proposition without any pretense of believing it. For instance, I might promise a friend that I will

defend some proposition against all challenges and counter-arguments, even while emphasizing

that I do not believe the proposition. How, then, can the non-restrictive dialectical perspective

explain why assertion requires apparent sincerity?

In §3, I noted that assertion occupies a dual role in reasoned discourse: through it, one

both undertakes and discharges dialectical commitments. So far, I have emphasized the first of

these two roles. As I will now argue, focusing on the second role helps defuse the present

objection to non-restrictive dialectical theories. Schematically: while apparent sincerity may not

be necessary for undertaking dialectical commitment, it is necessary for discharging dialectical

commitments; so assertion’s role in discharging dialectical commitment generates a constitutive

requirement of apparent sincerity.

I begin by exploring more closely the nature of “dialectical commitment.” In developing

the dialectical model, we should not depict reasoned discourse as a rhetorical exercise utterly

detached from the participants’ beliefs. By asserting a proposition, I make it my position, not

merely a position I agree to defend for the sake of argument. Assertoric commitment is not just a

commitment to providing arguments or to rebutting counter-arguments. As I will put it, assertion

involves a commitment to advocate for one’s position, i.e. to present it as worthy of belief.

Advocacy is not included in what the Defense and Default-Challenge Norms demand.

Specifically, it requires me to present my arguments as providing compelling reasons for

believing the asserted proposition. For instance, it is not enough merely to say, “Here are some

arguments that p, but I don’t myself find them plausible.” If I issue this disclaimer, I cease to

advocate for the asserted proposition. (A lawyer might tell the jury, “Here are some arguments

for my client’s innocence. I don’t find them convincing, but you should make up your own

mind.” The lawyer does not advocate for her client, no matter how impressive her arguments.)

How should we capture ‘advocacy’ through precise norms? I employ a fairly minimalist

formulation: advocacy involves presenting the premises of one’s argument as providing reason

for believing the asserted proposition. So construed, I think that advocacy underlies several

familiar versions of the dialectical model, including (Brandom, 1994). It is compatible with

various dialectical norms. For instance, we could supplement the Defense Norm with the proviso

that, when providing cogent arguments, one must advocate for one’s position. We could add an

analogous proviso to a norm demanding arguments that one’s interlocutor finds cogent, or that

the speaker finds cogent.

According to the dialectical model, assertion is individuated by its role in reasoned

discourse. Part of this role consists in the fact that speakers discharge dialectical commitments by

asserting propositions. So assertion, by its very nature, is a means for discharging dialectical

commitments. Clearly, I need not intend each assertoric performance as contributing to the

discharge of a dialectical commitment. I might assert p on a whim, without envisaging a broader

argumentative goal. Nevertheless, there is a clear sense in which even this whimsical assertion

contributes to the future putative discharge of dialectical commitments. Metaphorically: the

assertion requires no supplementation for me to cite p as a premise in future arguments. I need

merely re-affirm my prior assertion. This element of redundancy differentiates assertion from

speech acts like conjecturing p or putting p forward. If I conjecture p or put it forward, then I

cannot subsequently present it as affording reason for belief without fundamentally transforming

my standing towards it. If I assert p, no such transformation is required. By asserting p, I have

already presented it a potential premise in future arguments I may or may not offer.

As Brandom puts it, then, “[u]ttering a sentence with assertional force or significance is

putting it forward as a potential reason… Assertions are essentially fit to be reasons” (1994, p.

168). Part of what makes my speech act an assertion of p, rather than a conjecture or a mere

putting forward of p, is that I present p as providing reason for believing other propositions.

There may not be any specific proposition that I advance p as affording reason for believing. But

I present it as a reason to be invoked, when relevant, in future advocacy.

We may now offer the following argument that assertion constitutively requires a

pretense of sincerity:

(B) (i) By asserting a proposition, one presents it as providing reason for believing other

propositions.

(ii) One presents a proposition as providing reason for believing other propositions only

if one maintains some minimal pretense of believing it.

(iii) Hence, one asserts a proposition only if one maintains some minimal pretense of

believing it.

(i) and (ii) entail the desired conclusion (iii). The previous four paragraphs argued that (i) seems

plausible once we accept a suitable version of the dialectical model. (ii) seems plausible

whatever our conception of assertion. Suppose I utter p while indicating, perhaps through winks

or other body language, that I do not believe p. Then how can I present p as providing reason for

believing any other proposition q? I can advance it as a reason why some other person (including

my interlocutor) might believe q, but not as a reason why one should believe q. Lacking a

minimal show of credence in p, I cannot depict it as having any reason-giving force.

Revisiting an earlier example, suppose I promise a friend that I will vigorously defend p,

even while emphasizing to my friend that I do not believe p. Although I thereby undertake a kind

of dialectical commitment, I do not present p as affording reason for believing any other

proposition. Thus, my speech act does not contribute to the future putative discharge of

dialectical commitments. So it is not an assertion.

In motivating (i), we invoked structural features of reasoned discourse that any plausible

version of the non-restrictive dialectical perspective should accept. Moreover, (i) articulates a

constitutive requirement, not a constitutive norm, so there is no obvious entailment from (i) to

claims about assertion’s constitutive norms. Hence, (B) neither presupposes nor entails that

assertion falls under constitutive restrictive norms. In this way, our argument shows how a non-

restrictive dialectical account can explain why assertion requires apparent sincerity.

To buttress (B), imagine a dialectical practice that is “devil’s advocate all the way down.”

In this hypothetical practice, I can advance a proposition p, thereby incurring a commitment to

defend it. I defend p by advancing additional propositions as premises in an argument for p. My

interlocutor may in turn challenge those premises. However, neither my initial utterance of p nor

my subsequent defense of it need involve any pretense of sincerity. I merely advance and defend

propositions for the sake of argument, as when participants in our own dialectical practice

employ the following locutions: “Well, I’m not saying that I believe this myself, but some would

argue as follows,” “Proponents of p would respond to your counter-argument by saying q,” etc.

Could the hypothetical practice exist autonomously from normal assertoric practice? For

instance, must not participants in any stable, coherent dialectical practice have available a speech

act through which they at least pretend to believe claims about word-meanings or justificatory

relations between propositions? Leaving such questions aside, it seems clear that the hypothetical

practice differs dramatically from our own dialectical activity. The key difference is that, even if

participants in the hypothetical practice offer cogent arguments, they do not advance those

arguments as providing reasons for belief. In this sense, the practice hardly counts as a “game of

giving and asking for reasons.” Participants in the practice may provide arguments, but they do

not advocate for propositions. They undertake a fundamentally different kind of dialectical

commitment than that which we undertake through assertion. Apparently, then, any dialectical

practice remotely like ours requires a speech act through which one undertakes dialectical

commitments and throughout the performance of which one maintains some pretense of belief.

It is likely that reasoning similar to (B) can explain Davidson’s weakened version of

Unger’s Thesis, according to which someone who asserts p represents herself as believing p.

Specifically, we might replace (ii) with the following premise: one can present a proposition as

providing reason for believing other propositions only if one represents oneself as believing it.

What about the original version of Unger’s Thesis, based upon knowledge rather than belief? We

might invoke the much stronger premise (ii'): one can present a proposition as providing reason

for believing other propositions only if one represents oneself as knowing it. Although I do not

myself find (ii') particularly congenial, it strikes me as no less plausible than Unger’s Thesis.

Unger himself (1975, pp. 206-214) advocates a position somewhat reminiscent of (ii'): if X’s

reason for believing something is that p, then X knows that p. Williamson (2000, pp. 184-208)

advances a kindred doctrine: only known propositions are evidence for other propositions. Thus,

while the topic obviously requires further discussion, it hardly seems clear that Unger’s Thesis

poses insuperable difficulties for the non-restrictive dialectical perspective.

§6. Committing oneself to the truth of what one asserts

Peirce remarks that assertion involves an essentially normative ingredient: “This

ingredient, the assuming of responsibility, which is so prominent in solemn assertion, must be

present in every genuine assertion” (1934, p. 386). One can formulate Peirce’s insight in various

ways: a speaker commits himself to the proposition asserted; he takes responsibility for its truth;

he endorses the proposition; he stakes himself to the truth of the proposition; and so forth.

Many philosophers suggest that these Peircean formulations support the restrictive

model. For instance, Searle urges that “asserting commits the speaker to the truth of the

proposition asserted” (2001, p. 147). He infers that the Truth Norm is constitutive of assertion.

As he puts it, “[s]uch principles as ‘you ought to tell the truth,’ ‘you ought not to lie,’ or ‘you

ought to be consistent in your assertion’ are internal to the notion of assertion. You do not need

any external moral principle to have the relevant commitments. The commitment to truth is built

into the structure of the intentionality of the assertion” (2001, 181).

An obvious objection to Searle’s argument is that it rests upon the somewhat obscure

notion ‘committing oneself to the truth of a proposition.’ As MacFarlane (2005, p. 334) notes

while developing this objection, the most straightforward notion of commitment is commitment

to do something. I commit myself to attending a conference, to paying my Visa bill, and so on.

But an assertion that p is not, in general, a commitment to make it the case that p. In what sense,

then, do I ‘commit’ myself to the truth of a proposition that I assert?

One might propose that the relevant commitment is to believing the proposition, or

perhaps to judging that it is true. But what is it to undertake such a commitment? I can promise

to accumulate evidence and to survey it judiciously, but I can hardly promise what conclusions I

will draw. I can commit myself to self-inducing the belief that p, perhaps by swallowing a pill

with appropriate belief-altering properties. Obviously, this commitment is not an element of

normal assertion. On the other hand, if the putative commitment is to its already being the case

that I believe or judge p, then this revives the question of how to understand a commitment that

is not commitment to some course of action.

Another proposal is that, as Searle puts it, “[i]n making an assertion we take

responsibility for truth, sincerity, and evidence… These responsibilities are met only if the world

is such that the utterance is true, the speaker is sincere, and the speaker has evidence for the

assertion” (2001, p. 176). The relevant notion of “taking responsibility” is still fairly obscure. I

can take responsibility for doing something, such as watering my neighbor’s plants while he

vacations. I can take responsibility for having caused some proposition to be true. But what is it

to take responsibility for a proposition being true when I played no role in its truth? I can take

responsibility for compensating my interlocutor if what I assert turns out not to true. Yet this

obviously has no bearing upon typical assertoric performances. Another idea, explored by Alston

(2000, p. 54-63), is to construe ‘x takes responsibility for its being the case that p’ in terms of the

censure, blame, or reproach that x would or should encounter if it turns out that x did not believe

p. I will discuss these negative reactive attitudes in §7. Alston eventually rejects construals based

upon reactive attitudes, adopting instead the following analysis: “x takes responsibility for its

being the case that p in uttering s iff, in uttering s, x subjects his utterance to a rule that implies it

is permissible for x to utter s only if p” (p. 60). He argues that this rule-based analysis is more

explanatorily fundamental than one based upon negative reactive attitudes, since it explains the

source of those attitudes and also their pattern of distribution across various cases. However, the

rule-based analysis embodies something quite close to the restrictive model. Thus, as Alston

would doubtless acknowledge, it cannot subserve an argument for the restrictive model.

In ordinary conversation, we frequently adduce a speaker’s ‘conversational

commitments.’ We say that a speaker is committed to some proposition, or to some proposition

being true. We criticize a speaker who cannot defend his conversational commitments when

faced with challenges or counter-arguments. Thus, I think that Searle evokes fundamental

intuitions about assertion. But the non-restrictive dialectical perspective arguably accommodates

these Peircean intuitions just as well as the restrictive model. We can gloss ‘commitment to the

truth of a proposition’ as ‘commitment to defending (or advocating for) the truth of a

proposition.’ On this construal, a speaker who is ‘committed to a proposition’ or ‘committed to

the truth of a proposition’ is committed to defending the proposition within reasoned discourse.

Crucially, a commitment to defending the truth of what one says is not the same as a

commitment to speaking the truth in the first place.

It might seem that the proposed dialectical analysis provide a natural entrée for the

restrictive model. Commitment to defending the truth of what one says may not be literally the

same as commitment to speaking the truth. The former commitment may not even entail the

latter. But they seem to go together as a natural package. Why would I incur the former

commitment if I don’t also incur the latter?

The adversarial legal system provides an instructive, albeit imperfect, analogy. As

instantiated within the United States, this practice is governed by a complex set of constitutive

norms, some codified and others implicit in the common law tradition. Two lawyers advocate for

competing sides, with the verdict decided by a neutral third party. The lawyers are committed to

defending their positions as effectively as possible. Blatant failure to do so constitutes

malpractice and merits sanction. Yet Applbaum (2000) argues that advocacy within this system

falls under no global Honesty Norm. As Applbaum puts it, “over the course of representing a

client, a lawyer forms reasonable beliefs… about the various factual propositions at issue in the

trial, and… zealous advocacy often requires her to attempt to persuade the jury to believe the

opposite” (p. 106). The relevant ‘reasonable beliefs’ sometimes rise to the level of knowledge.

Legal norms do not allow lawyers to say just anything. Complex rules of evidence forbid certain

especially egregious deceptions. But no generalized Honesty Norm prevails. Thus, there is a

clear sense in which legal advocates are not, in general, committed to speaking the truth.

From the non-restrictive dialectical perspective, assertion exhibits many similarities with

the adversarial legal system. By asserting a proposition, I commit myself to advocating for it, just

as a lawyer who takes a case commits herself to advocating for her client’s position. There is no

constitutive commitment to advocate only for positions one believes, only a commitment to

advocate for whatever position one chooses. In neither case, then, does commitment to defending

the truth of what one says entrain commitment to speaking the truth in first place. Of course,

there are many disanalogies between assertion and the adversarial system: legal practice involves

explicitly codified norms directly sustained by the threat of punitive sanctions; there are

supposedly neutral adjudicators distinct from the advocates; part of practice’s rationale is the

need to preserve individual rights; and so on. But the non-restrictive dialectical approach holds

that these disanalogies conceal a more fundamental homology in normative structure.

This picture of assertion strikes many philosophers as absurd. It is admittedly a bit

extreme. So far, though, we have found no cogent argument against it.

§7. Intuitions of speaker-defectiveness

Philosophers often motivate the restrictive model by adducing pre-theoretic intuitions of

defectiveness. We may divide the relevant intuitions into two categories: intuitions that the

speaker is somehow blameworthy; and intuitions that, whether or not the speaker is culpable, her

assertion itself is defective.

I will describe these intuitions respectively as speaker-defect

intuitions and assertion-defect intuitions. A liar might elicit speaker-defect intuitions. A speaker

who asserts a false proposition she has excellent reason to believe might elicit assertion-defect

intuitions. I discuss speaker-defect intuitions in this section and assertion-defect intuitions in §8.

If we find that a speaker lied, we may blame or censure her. The restrictive model can

explain these negative reactive attitudes as registering that the speaker deliberately violated some

constitutive assertoric norm. But do the negative reactive attitudes reflect anything special about

assertion? Or do they reflect general norms that encompass both assertion and many other

actions? A useful technique here is to elicit comparative judgments regarding analogous cases of

assertoric and non-assertoric communication. Non-assertoric counterparts function somewhat

like controls in a scientific experiment.

Consider the following examples, several of which derive from two famous cases

discussed by Kant: the “murderer at the door,” and the man who packs his luggage to create the

misleading impression he is embarking on a trip.

1(a): John is helping an innocent victim evade a killer who also is John’s neighbor. The victim is

hiding in John’s car. To fool the killer, who does not know that John is harboring the victim but

who is standing far down the street, John goes to his car with a suitcase and shouts to the killer,

“I’ll be out of town for a few days,” then drives away. He does not intend to leave town.

1(b): Similar to 1(a), except that this time John says nothing to the neighbor. He packs a heavy

suitcase and carries it to his car. He waves to his neighbor, and he points silently to the suitcase

while loading it in the car. He does this to create the same misleading impression as in 1(a).

2(a): While preparing to drive to a marathon poker game at a nearby location, John shouts to his

nosy but harmless neighbor far down the street, “I’ll be out of town for a few days.” He lies

because he does not want his neighbor to know why he will be gone.

2(b): Stands to 2(a) as 1(b) stands to 1(a).

3(a): Iago tells Othello that Desdemona betrayed him with Cassio.

3(b): Iago shows Othello doctored photographs of Desdemona betraying him with Cassio.

In each pair of examples, an agent conveys roughly the same proposition. But he conveys it

assertorically in (a) and non-assertorically in (b). If the Honesty, Truth, Warrant, or Knowledge

Norm is constitutive of assertion, then (a) exhibits a specifically linguistic defect beyond

whatever defects it shares with (b). Unlike (b), (a) violates the constitutive norms of assertion.

Does the deceiver in (a) exhibit a qualitatively different type of defect than in (b)? I do

not think so. Our relevant intuitions within each pair of examples strike me as commensurate.

We view 1(a-b) as commendable, 2(a-b) as involving at most a fairly minor infraction, and 3(a-

b) as involving a serious infraction. Assertoric communication is quite different than non-

assertoric communication, so we might cite many disanalogies between each (a) scenario and the

corresponding (b) scenario. But it is hard to see how these disanalogies secure any intuitive, pre-

theoretic respect in which (a) elicits a fundamentally different censorious reaction than (b). Thus,

in explaining (a)’s blameworthiness, or lack thereof, there is no need to invoke specialized

assertoric norms. We can invoke general evaluative standards that apply both to assertion and to

many other actions. When assessing the deceiver for praise or blame, the fact that he employed

assertion as his mode of deception seems negligibly significant.

An imposing philosophical tradition, famously represented by Aquinas and Kant, insists

that direct lies are somehow worse than other forms of deception.

Bernard Williams, who

accuses Aquinas and Kant of “fetishizing assertion,” even so concedes that under certain

circumstances “there is something peculiarly odious or insulting about a lie as contrasted with

other forms of deceit” (2002, p. 118). The most popular illustration is conversational

implicature. As Augustine observed, there are many contexts in which misleading conversational

implicatures seem far less objectionable than outright lies. On this basis, Adler (1997) argues that

assertion, unlike conversational implicature, falls under a constitutive norm of truthfulness. As

another example, imagine a variant of scenario 2(b) in which John carries his luggage to his car

but pretends not to notice his neighbor observing him. This scenario, which is similar to how

Kant himself intended the luggage example, seems markedly less “odious and insulting” than a

direct lie. At the very least, such examples suggest that a lie’s defects outstrip its role in

disseminating propositions the speaker believes false. A lie is not bad simply because it deceives

or because it is intended to deceive.

Nevertheless, we should not conclude that lies are in general more deplorable than other

forms of deceit. That conclusion seems plausible only if we focus on a narrow range of

unrepresentative examples. For instance, the deceived neighbor in Kant’s version of the luggage

example does not know that John seeks to influence her beliefs. The example therefore involves

nothing resembling common knowledge of communicative intentions and expectations. Once we

incorporate this common knowledge, as in 2(b), John’s deceit becomes far more objectionable.

Now, John not only dupes his neighbor, he also abuses her trust. Given this abuse of trust, 2(b)

seems no less “odious and insulting” than 2(a).

Misleading conversational implicature is likewise an unrepresentative example.

Conversational implicature inherently lends itself to craftiness. It is a way of conveying

propositions without committing oneself to them. If I conversationally implicate p rather than

asserting p, it is natural for my interlocutors to wonder why. Why do I beat around the bush,

rather than just come out and say what I mean? In many contexts, then, the decision to implicate

p rather than assert p signals caveat auditor. As Adler himself observes, “the choice of a

deception rather than a lie can itself convey a message of moral effort and struggle” (1997, p.

452). Adler’s observation helps explain why misleading conversational implicatures often seem

less offensive than outright lies. Unlike the liar, a speaker who misleads through conversational

implicature provides an implicit warning, through his choice of communicative strategy, that we

should parse his message with extreme care. If his deception does not seem so bad, that is partly

because his mode of deception already broadcasts his guile.

In contrast, the modes of deception employed in scenarios 1-3(b) raise no red flags. We

may construe each scenario so that it seems quite natural for the deceiver to communicate his

message non-assertorically: in 1-2(b), because John’s neighbor is quite far away; and in 3(b),

because the photographs speak for themselves, requiring no verbal supplementation. Once

construed in this way, 2-3(b) evince intuitive defects comparable to 2-3(a). Thus, excessive focus

upon conversational implicature, at the expense of more suitable examples, encourages the

mistaken impression that assertoric deceit is peculiarly repugnant.

My goal is not to analyze why lying is wrong or how it relates to other forms of

deception. As the ethical literature on mendacity demonstrates, a complete analysis would

require sophisticated casuistical maneuvers applied to a wide range of cases. A good account

might invoke some of the following familiar ideas: deception often yields bad consequences for

the deceived; the deceiver abuses a bond of trust with the deceived; the deceiver acts on a maxim

that is not universalizable; the deceiver treats the deceived as a means, not as an end; the

deceiver violates our transpersonal rational interest in the pursuit of truth; the deceiver

undermines our trust in one another and thereby diminishes our ability to communicate; and so

on. My point is that, despite these complexities, we lack any antecedent reason to expect that a

good account will stigmatize assertoric deception as uniquely blameworthy when compared to

suitable varieties of non-assertoric deceit. A special pejorative stance towards assertoric

deception may be a consequence of a satisfying theory, but it is not a pre-theoretic constraint.

I turn now to cases where the speaker believes the asserted proposition but lacks

“sufficient” grounds for asserting it. Such cases often elicit the intuition that the speaker, while

not as culpable as a liar, behaves irresponsibly. The restrictive model can explain this intuition

by saying that the speaker exerts insufficient effort to obey constitutive assertoric norms.

I begin with an example made famous by Williamson. The lottery results will be

announced in a few minutes, and my friend is waiting to hear whether her ticket won. She

purchased the ticket on a lark, so she is not very invested in the outcome. I say, “Your ticket did

not win.” Call the proposition asserted “the lottery proposition.” My grounds for believing the

lottery proposition are purely probabilistic. I do not have any inside information about the

lottery. According to Williamson, I do not know the lottery proposition, even though I am

justified in believing it. Williamson also claims that “you will still be entitled to feel some

resentment when you later discover the merely probabilistic grounds for my assertion” (p. 246).

Williamson argues that we should explain this intuitive reaction through the Knowledge Norm.

Given the details of the lottery example, it is difficult to construct a clean non-assertoric

counterpart akin to 1-3(b). But we can study some related cases:

4(a): John’s company is planning a weekend retreat in the country. The company is about to

announce the result of a lottery, conducted last night. The lottery’s winner, if any, will be

excused from the retreat. Everyone else must attend. John is leaving for the office to learn the

lottery’s result. If his ticket won, he is determined to stay in town. If not, he is determined to go

on the retreat. He knows his ticket has very low odds of winning (let them be as low as you

please). He brings luggage with him so that he can leave directly for the retreat if necessary. As

he enters his car, he shouts to his neighbor, “I’ll be out of town for a few days.”

4(b): Similar to 4(a), except that John says nothing. Instead, he packs a heavy suitcase, waves at

his neighbor, and points silently at the suitcase while loading it into his car. He does this to

communicate that he will be out of town for a few days.

It seems fairly clear that John’s basis for believing he will leave town is probabilistic. By

Williamson’s lights, John does not know he will leave town. Suppose that, in both 4(a) and 4(b),

John’s neighbor eventually learns about the lottery and its significance. Williamson’s account

predicts that the assertoric case merits more resentment than the non-assertoric one. I think it far

from clear that pre-theoretic intuition validates this prediction. As long as John’s odds of

winning the lottery are low enough, I am not sure that 4(a) elicits any resentment at all. Even if it

does, I do not think it seems any worse than 4(b). Both cases strike us as, at worst, mildly

irresponsible but hardly grounds for strident criticism.

We can imagine a series of variants upon 4(a) and 4(b), in which we alter the nature and

strength of John’s warrant for believing that he will be out of town for a few days. In each pair of

cases, I conjecture that John elicits no more resentment in (a) than in (b).

Generally, the literature sparked by Williamson’s discussion has debated which

restrictive assertoric norm best explains the lottery example, not whether we should explain the

example through a restrictive norm. Weiner (2005) argues that we should explain the lottery

example through the Truth Norm and Gricean implicature, while Kvanvig (2009) favors an

account based on the Warrant Norm. Such treatments, like Williamson’s, typically cite speaker-

defect intuitions without paying even cursory attention to non-assertoric counterparts. In my

view, this is somewhat like trying to explain why property X causes cancer before checking that

subjects with X acquire cancer more frequently than subjects lacking X. Before documenting

intuitive discrepancies with comparable non-assertoric cases, we should not assume that negative

reactions elicited by assertoric performances reflect anything special about assertion.

The parity between 1-4(a) and 1-4(b) suggests that speaker-defect intuitions provide little

support for the restrictive model. Any resentment evoked by 1-4(a) reflects not specialized

assertoric norms but general norms governing belief-transmission through both linguistic and

non-linguistic means. Admittedly, I have not formulated those general norms, nor is it obvious

how to do so. But there is no evident reason why the patterns of blame and praise surrounding 1-

4(a) require a special framework of restrictive constitutive norms. On the contrary, that

framework predicts disparities that find no echo in pre-theoretic intuition.

In response to my argument, proponents of the restrictive model might propose that 1-

4(b) themselves fall under constitutive restrictive norms. On this proposal, speaker-defect

intuitions register no disparity between (a) and (b) because they both violate constitutive norms.

The challenge facing this proposal is to explain why 1-4(b) involve a practice with the

features delineated in §2: internal evaluative standards codified by constitutive norms. For

instance, in 1-2(b) and 4(b), John communicates that he will be out of town for a few days by

exploiting his neighbor’s mastery of folk psychology, not mastery of some pre-existing practice.

Given John’s conduct, it is common knowledge between John and his neighbor that he wants the

neighbor to notice his suitcase. The most natural explanation for why John is loading a heavy

suitcase into his car is that he is embarking on a trip that will last several days. So it is reasonable

for the neighbor to infer that John is trying to instill the belief that he is embarking on a trip that

will last several days. Since the neighbor believes from prior interactions with John that he is

honest and reliable, the neighbor concludes that John is embarking on a trip that will last several

days. This explanation does not depend in any obvious way upon constitutive norms.

§8. Intuitions of assertion-defectiveness

Assertion-defect intuitions record that the assertion itself is defective, even if the speaker

is not culpable. As Alston puts it, “the utterance is wrong, incorrect, or out of order, even if the

epistemological conditions for blameworthiness are not satisfied… [I]t was the wrong thing to

say, whatever the correct judgment as to the speaker’s fault… [T]here was something

fundamentally amiss with the utterance” (2000, p. 57, 60). The point here is not just that some

assertion is, say, false. That is something all sides can acknowledge. The point is that describing

an assertion as false constitutes a criticism of it. By asserting a falsehood, the speaker fails to

meet some relevant evaluative standard, no matter how hard she tried to reach it. As Alston puts

it, “we must distinguish the perception of an out-of-order sentence utterance from the judgment

that the assertion, or what is asserted, is false” (2000, p. 267). Similarly, one might claim that an

assertion seems ‘amiss’ if the speaker does not believe what she asserts, or does not believe it

with sufficient warrant, or does not know it. The restrictive model can explain these intuitions as

registering that the speaker, perhaps unwittingly, violated some constitutive assertoric norm.

A basic difficulty here is that assertion-defect intuitions are much less robust than

speaker-defect intuitions. Suppose that, mistakenly believing it is raining, I assert that it is

raining. We can all agree that my assertion is false. But is it clear that this verdict constitutes a

criticism? Do we really have a ‘perception’ that the utterance is ‘out-of-order,’ over and above

our recognition that it is false? Matters become even less clear for 1(a), in which the assertion’s

falsity and the speaker’s dishonesty might well seem the basis, not for criticism, but for

unqualified acclaim. Of course, advocates of the Truth or Honesty Norms insist that 1(a) seems

intuitively defective qua assertion, despite any countervailing moral or prudential merits. But the

putative ‘intuition’ here seems perilously close to an intuition that the restrictive model is

correct. Opponents of the restrictive model, such as Davidson, claim not to share such intuitions.

How can we decide between the restrictive and non-restrictive perspectives if we cannot

agree upon the data to be explained? Do we face a Kuhnian stalemate in which rival theorists

adduce divergent theory-laden intuitions? To some extent, perhaps. Before giving up, though, I

will examine a few representative arguments advanced by proponents of the restrictive model. I

will urge that those arguments assign assertion-defect intuitions a problematic role.

Consider the following passage from Searle (2001, p. 184):

Suppose I am not lying, but am genuinely mistaken. I sincerely said it is raining, but all

the same it is not raining. In such a case there still is something wrong with my speech

act, namely, it is false. But why is that wrong? After all, for every true proposition there

is a false one. It is wrong because the aim of a statement is to be true, and this one fails,

because it is false. When I make a statement I commit myself to its truth, and here my

mistake makes me fail in my commitment.

This passage moves in a tight circle between theory and data, rendering the line between premise

and conclusion difficult to discern. Apparently, Searle seeks to elicit assertoric-defect intuitions

by invoking two ideas: first, assertion involves a commitment to the truth of what one says;

second, the aim of assertion is truth. I critiqued the first idea in §6. The second idea seems too

close to the restrictive model to provide a non-circular argument for it. Setting these criticisms

aside, it seems clear that, rather than motivating the restrictive model by citing assertion-defect

intuitions, Searle motivates the intuitions by citing putative normative features of assertion. But

then Searle can hardly count the intuitions as pre-theoretic data upon which we should all agree.

Unlike Searle, Alston emphasizes that assertion-defect intuitions are elusive. His solution

is to examine first language acquisition (2000, p. 267):

We seek to get across to the small child that he is not to say “A rabbit is outside” unless

there is a rabbit outside… We try to get across to him when it is and when it is not all

right to for him to utter a certain sentence… The details of first language acquisition are

enormously complicated. But one thing that is crucially involved is the disposition on the

part of teachers to spot cases in which the learner utters a sentence in violation of an I-

rule and apply appropriate correction,

where ‘I-rule’ is Alston’s term for ‘constitutive norm of some speech act.’ Alston concludes that

ordinary practice enshrines a ‘perception’ that false assertions are ‘out-of-order,’ a perception

made manifest when we inculcate children into linguistic practice.

Alston observes that we encourage children to strive for true assertions. In fact, few

parents teach their children that false assertions are never ‘all right,’ since lies are sometimes

appropriate. Ignoring this quibble, Alston’s observation still does not support the restrictive

model. As I argued in §5, the non-restrictive dialectical perspective can explain why assertion

requires a pretense of sincerity. Since asserting p requires a pretense that one believes p,

asserting p is liable to induce others to believe p. Since moral norms require one, ceteris paribus,

to avoid instilling false beliefs in others, they require one, ceteribus paribus, to strive to assert

the truth. Hence, encouraging children to strive for true assertions might just be one chapter in

their moral education. So Alston’s observation does not suggest that a false assertion exhibits

any defect beyond whatever moral, epistemic, cognitive, or rational defects may have led to it.

Alston also observes that, when we discover a child has asserted a falsehood, we ‘apply

appropriate correction.’ Yet how do we apply this ‘appropriate correction’? Prototypically, we

confront the child with evidence that he asserted a false proposition and require him to concede,

at least tacitly, that he was wrong. The non-restrictive dialectical perspective can explain this

interaction as an instance of reasoned discourse, in which we try to induce the child to obey the

Retraction Norm. The fact that we expect someone to retract an assertion does not suggest that

the original assertion was defective, only that refusal to retract it would be defective.

We must distinguish two theses. The first thesis is that, when a speaker’s interlocutors

produce decisive counter-arguments against her assertion, then she must retract it. The second

thesis is that the original assertion was itself somehow defective. Our intuitions strongly support

the first thesis: if a speaker simply will not listen to reason (“That’s my story and I’m sticking to

it”) then we find her linguistic conduct highly defective. But that suggests only something like

the Retraction Norm. It does not follow that false assertions, or even assertions known by the

speaker to be false, violate some constitutive assertoric norm.

Following Searle and Alston, I have focused upon cases in which a speaker unwittingly

asserts a false proposition. What about cases where the speaker believes a true proposition but

apparently lacks ‘sufficient’ grounds for asserting it? Consider again the lottery proposition.

Although Williamson recognizes the distinction between intuitions of speaker- and assertion-

defectiveness (p. 261), he often blurs it, as in the following passage: “Intuitively, my grounds are

quite inadequate for that outright unqualified assertion… You will still be entitled to feel some

resentment when you later discover the merely probabilistic grounds for my assertion. I was

representing myself to you as having an authority to make the flat-out assertion which in reality I

lacked’” (p. 246). The first and third sentences of this passage strike me as ambiguous between

speaker- and assertion-defectiveness, while the second sentence evokes speaker-defectiveness.

Taken as a whole, the passage confers spurious credibility upon assertion-defect intuitions by

eliding the distinction with speaker-defect intuitions. This is no isolated expository anomaly.

Discussions of the lottery example frequently either elide the distinction or else focus exclusively

upon speaker-defect intuitions ((Weiner, 2005) illustrates the latter option).

Another worry centers again upon the crucial sentence: “You will still be entitled to feel

some resentment when you later discover the merely probabilistic grounds for my assertion.”

This sentence naturally invites us to consider an extended scenario in which I fail to defend the

lottery proposition against challenges. It therefore elicits intuitions about how I should behave

after asserting the lottery proposition, which are easily confused with intuitions about the

assertion itself. For instance, suppose friend asks, “Why do you think my ticket didn’t win?

What’s your evidence?”. I might respond by citing the low probability that the ticket won. But

my friend could retort: “I know all about the probabilities. What I want from you is an argument

that my ticket did not win, not an argument that it has a low chance of winning.” And to this I

would lack any adequate reply. I would fail to vindicate my position. Williamson regards this

failure as evidence of linguistic defectiveness. But one might instead regard the failure,

assuming it is accompanied by a refusal to retract, as itself the sole linguistic defect. It is not

evidence for a further transgression of ‘exceeding one’s evidential authority,’ a transgression that

already occurred with the original assertion. We must distinguish one undeniable intuition, that I

should retract my assertion when I prove unable to defend it, from a further, much more

debatable intuition that the original assertion was defective.

In general, it seems far from clear that we experience robust assertion-defect intuitions

for the lottery example, or for other examples where the speaker asserts a proposition without

‘sufficient’ grounds. We may have an intuition that the speaker behaves irresponsibly, and we

may have an intuition that the speaker must retract his assertion after failing to defend it. I argued

in the previous section that we should explain the former intuition through general moral and

social factors, not through constitutive assertoric norms. The latter intuition suggests only

something resembling the Defense or Default-Challenge Norm. I submit that we can explain any

non-controversial intuitions elicited by the lottery example, and other similar examples, through

general moral and social norms combined with non-restrictive dialectical norms.

My analysis of the lottery example is consistent with almost any epistemological

treatment. In particular, I take no stand regarding whether we know the lottery proposition or

even whether we are justified in believing it. My analysis cites only an inability to provide an

adequate argument supporting that proposition. As Alston (1989) and Audi (1993) emphasize,

the ability to produce arguments for some proposition is quite distinct from the state of being

justified in believing the proposition, let alone the state of knowing it.

Does my treatment of the lottery case presuppose that probabilistic arguments never

discharge dialectical commitments? That presupposition would be tendentious. For instance,

John might defend the assertion that he has some disease by citing the results of a blood test. The

test has an extremely slight chance of false positives, so John’s defense appears probabilistic. Yet

the defense will strike many as perfectly legitimate. Why might we deem probabilistic

argumentation acceptable in this scenario but not in the lottery example?

An adequate answer to this question would require a thorough analysis of probabilistic

justification, a vexed topic in contemporary epistemology.

Even lacking such an analysis, there

is clearly something special about the lottery scenario that sets it apart from more routine cases

like the blood test, as reflected by our unwillingness to say that one knows the lottery proposition

but not that one can know one has a disease based on a blood test. Abductive considerations may

help differentiate the two cases. The best explanation for why John’s blood test is positive is that

he has the disease; so a defense based on the blood test can appeal not just to ‘bare’ probabilities

but to abduction. Abductive considerations are irrelevant to the lottery proposition, which is a

prediction based upon good odds, not an explanation of anything.

Thus, one might urge that

probabilistic arguments are legitimate when they potentially admit abductive supplementation.

Whether or not this maneuver generalizes to all other intuitively acceptable probabilistic

arguments, it shows that I am not committed to any global denunciation of probabilistic

argumentation in general. My analysis assumes only that, given the specifics of the lottery

scenario, citing probabilities does not constitute an adequate defense of the lottery proposition.

Ultimately, I simply reject the putative assertion-defect intuitions that Alston, Searle, and

Williamson seek to explain. However, our discussion suggests that assertion-defect intuitions are

non-optimal candidates for pre-theoretic linguistic data. Although suggestive, they are theory-

laden and labile. They are also easily confused with other quite different intuitions. This does not

render them useless. An adequate theory of assertion must catalogue and account for them. But it

must do so with greater circumspection than proponents of the restrictive model often exercise.

§9. The non-restrictive dialectical perspective

Evidently, it is more difficult than one might suppose to provide a compelling argument

for the restrictive model. Although the literature contains many arguments besides those I have

discussed, I believe that my discussion casts favorable light upon the non-restrictive dialectical

perspective.

That perspective is much more appealing than one would suspect based on how few

contemporary philosophers advocate it. We should investigate it in more detail. At the very least,

such investigations may bring into sharper relief the strengths and weaknesses of the restrictive

model.

I am greatly indebted to helpful feedback from Jonathan Adler, Anthony Corsentino, Kevin Falvey, Michael

Glanzberg, Warren Goldfarb, Elizabeth Harman, Richard Heck, Louis-Philippe Hodgson, Charles Parsons, and

James Pryor. I received many useful comments when I presented earlier versions of this paper at a UCSD

Philosophy Colloquium, especially from Agustín Rayo and Wayne Martin, and at the Pacific APA 2005. I am

grateful to Gary Ebbs, my commentator on the latter occasion, for his criticisms. Finally, I thank an anonymous

referee for this journal for unusually generous and helpful comments.

Does this approach entail that, if a speaker cannot satisfactorily defend p against a recalcitrant interlocutor, then the

speaker is not justified in believing p? No. My position concerns norms of assertion, not norms of belief. There are

no easy inferences from the former to the latter. As Alston (1989) and Audi (1993) argue, we must sharply

distinguish the ability to provide arguments for p from the state of being justified in believing p.

One complication is that “presenting x as y” is an opaque context. Thus, even if we agree with Williamson that

knowledge=evidence, we cannot immediately infer that one presents a proposition as evidence for other propositions

only if one represents oneself as knowing it; similarly for any putative inference from Unger’s position to (ii').

Nevertheless, it may be that the arguments offered by Unger and Williamson, when suitably modified, support

something sufficiently close to (ii').

Roughly speaking, speaker- versus assertion-defect intuitions correspond respectively to what DeRose (2002) calls

“secondary” versus “primary” violations of constitutive assertoric norms and what Alston (2000) calls shirking of

one’s “subjective” versus “objective” obligations.

(Scanlon, 1990, p. 202-3) employs an analogous argumentative strategy, vis-à-vis promising rather than asserting.

(Scanlon, 1999, p. 317-322) explains the defectiveness of lying in terms of generic moral considerations that also

apply to non-linguistic communication. See also the discussion of promising in (Cavell, 1979, p. 298).

For an overview of this tradition, with citations, see (MacIntyre, 1995).

For discussion, with citations to the literature, see (Nelkin, 2000). Nelkin argues that bare probabilistic grounds are

not sufficient for rational belief. She argues that beliefs based upon non-deductive grounds can nevertheless be

rational when those grounds incorporate some abductive element.

Cf. (Nelkin, 2000, p. 404-408).

Alston (2000, p. 78), DeRose (2002), (Williamson, 2000, p. 253), and many others argue that Moore’s paradox

supports the restrictive model. In contrast, Shoemaker (1996, p. 74-96) and Moran (2001) argue that Moore’s

paradox is fundamentally a phenomenon at the level of cognition, rather than language. The locus of deviancy lies

not with the assertions “p, but I don’t know that p” or “p, but I don’t believe that p,” but with the corresponding

beliefs. As Shoemaker argues, we should first explain why there is something deviant about believing these

propositions. Once we do so, it is unlikely that we will require substantial philosophical machinery to explain why

there is something deviant about asserting them.

Another important observation is that one can appropriately answer any assertion by saying, “How do you

know?”. Unger deploys this observation to motivate Unger’s Thesis, and Williamson deploys it to motivate the

constitutive status of the Knowledge Norm. For critical discussion, see (Kvanvig, 2009).

Works Cited

Adler, J. 1997. Lying, deceiving, or falsely implicating. Journal of Philosophy 94, 435-452.

Alston, W. 1989. Epistemic Justification: Essays in the Theory of Knowledge. Ithaca: Cornell University Press.

---. 2000. Illocutionary Acts and Sentence Meaning. Ithaca: Cornell University Press.

Applbaum, A. 2000. Ethics for Adversaries. Princeton: Princeton University Press.

Audi, R. 1993. The Structure of Justification. Cambridge: Cambridge University Press.

Brandom, R. 1994. Making it Explicit. Cambridge: Harvard University Press.

Cavell, S. 1979. The Claim of Reason. Oxford: Oxford University Press.

Davidson, D. 1984. Inquiries into Truth and Interpretation. Cambridge: Cambridge University Press.

---. 2005. Truth and Predication. Cambridge: Harvard University Press.

DeRose, K. 2002. Assertion, knowledge, and context. Philosophical Review 111, 167-203.

Douven, I. 2006. Assertion, knowledge, and rational credibility. Philosophical Review 115: 449-485.

Hamblin, C. L. 1971. Mathematical models of dialogue. Theoria 37, 130-155.

Kvanvig, J. 2009. Assertion, knowledge, and lotteries. Williamson on Knowledge, ed. D. Pritchard and P.

Greenough. Oxford: Oxford University Press.

Leite, A. 2005. A localist solution to the regress of justification. Australasian Journal of Philosophy 83, 395-421.

MacFarlane, J. 2003. Future contingents and relative truth. The Philosophical Quarterly 53, 250-65.

---. 2005. Making sense of relative truth. Proceedings of the Aristotelian Society 105, 321-39.

MacIntyre, A. 1995. Truthfulness, lies, and moral philosophers. The Tanner Lectures on Human Values, Volume

XVI, ed. G. B. Peterson. Salt Lake City: Utah University Press.

Moran, R. 2001. Authority and Estrangement. Princeton: Princeton University Press.

Nelkin, D. 2000. The lottery paradox, knowledge, and rationality. Philosophical Review 109, 373-409.

Pierce, C. S. 1934. Collected Papers, vol. V. Cambridge: Harvard University Press.

Rawls, J. 1955. Two concepts of rules. Philosophical Review 63, 3-32.

---. 1971. A Theory of Justice. Cambridge: Harvard University Press.

Rescorla, M. 2007. A linguistic reason for truthfulness. In Truth and Speech Acts, ed. D. Greimann and G. Siegwart.

New York: Routledge.

---. 2009a. Epistemic and dialectical regress. Australasian Journal of Philosophy 87, 43-60.

---. 2009b. Shifting the Burden of Proof?. The Philosophical Quarterly 59, 86-109.

Rumfitt, I. 1995. Truth-conditions and communication. Mind 104, 827-862.

Scanlon, T. M. 1990. Promises and practices. Philosophy and Public Affairs 19, 199-226.

---. 1999. What We Owe to Each Other. Cambridge: Harvard University Press.

Searle, J. 1969. Speech Acts. Cambridge: Cambridge University Press.

---. 2001. Rationality in Action. Cambridge: MIT Press, 2001.

Sellars, W. 1963. Some reflections on language games. Rpt. in Sellars’s Science, Perception, and Reality. New

York: Routledge and Kegan Paul.

Shoemaker, S. 1996. The First-Person Perspective and Other Essays. Cambridge: Cambridge University Press.

Thomson, J. J. 1990. The Realm of Rights. Cambridge: Harvard University Press.

Toulmin, S. 1958. The Uses of Argument. Cambridge: Cambridge University Press.

Unger, P. 1975. Ignorance. Oxford: Clarendon Press.

Watson, G. 2004. Asserting and promising. Philosophical Studies 117, 55-77.

Weiner, M. 2005. Must we know what we say? Philosophical Review 114, 227-251.

Williams, B. 2002. Truth and Truthfulness. Princeton: Princeton University Press.

Williamson, T. 2000. Knowledge and its Limits. Oxford: Oxford University Press, 2000.