P, but you don’t know that P

Abstract

Unlike first-person Moorean sentences, it’s not always awkward to assert, “p, but you don’t know that p.” This can seem puzzling: after all, one can never get one’s audience to know the asserted content by speaking thus. Nevertheless, such assertions can be conversationally useful, for instance, by helping speaker and addressee agree on where to disagree. I will argue that such assertions also make trouble for the growing family of views about the norm of assertion that what licenses proper assertion is not the initiating epistemic position of the speaker but the (potential) resulting epistemic position of the audience.

“There’s something you must remember: You’re braver than you believe,

and stronger than you seem, and smarter than you think.”

—Christopher Robin, to Winnie the Pooh.

1 Introduction

Assertion is our most straightforward practice for informing others about the world. To assert that p is (at least) to say a sentence that expresses p (in the given context) in the right sort of way. As so often in philosophy, “the right sort of way” turns out to be terribly difficult to characterize. But since Williamson (2000), some philosophers have thought that part of the story involves assertion being subject to a special, epistemic norm—at least a norm that applies to assertions that are straight-faced and unhedged, spoken in ordinary contexts.¹ This paper, which belongs to that tradition, argues that second-person Moorean assertions make trouble for the growing family of views about the norm of assertion: the view that what licenses proper assertion is not the epistemic position of the speaker but of the audience.

But first, why think that the norms of assertion reference audiences in the first place? In direct challenge to the knowledge norm, and other norms requiring the speaker to believe what they assert, Jennifer Lackey proposes that in cases of selfless assertion, the asserter can help their audience come to believe what they themselves cannot. Consider Sebastian, the distraught doctor:

DISTRAUGHT DOCTOR: Sebastian is [a] ...pediatrician [who] recognizes ...that all of the scientific evidence shows that there is absolutely no connection between vaccines and autism. However, shortly after his ...daughter received one of her vaccines, ...she was soon diagnosed with autism. …[T]he grief and exhaustion brought on by his daughter’s recent diagnosis cause him to abandon his previously deeply-held beliefs regarding vaccines. …[W]hile performing a well-baby checkup..., the child’s parents ask him about ...the rumors surrounding vaccines and autism. Recognizing both that the current doubt he has towards vaccines was probably brought about through [his] emotional trauma ...and that he has an obligation to his patients, ...Sebastian asserts, “There is no connection between vaccines and autism” …[although it’s false that] Sebastian himself believes or knows this proposition. (Lackey, 2007, pp. 598–599)

Lackey concludes that “it is a mistake to require proper assertion to pass through the doxastic states of the asserter” (Lackey, 2007, p. 600). Doffing the belief requirement on assertion allows speakers like Sebastian to help their audiences learn despite the epistemic limitations of the speaker. What Lackey recognizes is that assertion is, to some extent, for audiences—at least in the normal case.² Lackey thus proposes a norm of assertion that is audience-accommodating. It allows speakers to assert what they don’t believe (in certain cases) if doing so promotes the epistemic well-being of the audience:

RTB-Norm: One should assert that p only if it is reasonable for one to believe that p. (Lackey, 2007, p. 608)³

Other authors, too, have crafted norms with the intent to capture her audience-accommodating insight:

SRN-Norm: One may assert that p only if one has supportive reasons for p. (McKinnon, 2015, p. 4)⁴

PtK-Norm: One should assert that p only if one is in a position to know that p. (Willard-Kyle, 2020)

Unsurprisingly, disagreement persists: some philosophers think that our judgments about Lackey’s selfless assertions are misleading (e.g. Kvanvig, 2011, p. 235) or that her cases can be explained without giving up on belief-entailing norms (Sosa, 2011, p. 47; Turri, 2015). But in this paper, we’ll be interested not in the objection that Lackey’s audience-accommodation goes too far but that it doesn’t go far enough.⁵ Note that the audience-accommodating norms just considered still maintain that proper assertion depends on the epistemic position of the speaker. One should assert only if it is reasonable for the speaker to believe, if the speaker has supportive reasons. But once we’ve reoriented ourselves around the idea that assertion is (typically) aimed at an audience, we may start to wonder whether it isn’t the epistemic position of the audience that really licenses proper assertion.⁶

This general strategy has been recently advanced by Manuel García-Carpintero, Charlie Pelling, and Edward Hinchman. Their norms are not just audience-accommodating but audience-centric:

TKR: One must ((assert p) only if one’s audience comes thereby to be in a position to know p) (García-Carpintero, 2004, p. 134; 2020, p. 270).

PKA: One’s assertion that p is proper only if it is fit to give a hearer knowledge that p (Pelling, 2013, p. 297).⁷

PTW: Assert [that p] only when you could provide testimonial warrant [that p] to a potential addressee (Hinchman, 2020, p. 556).

These norms perform surprisingly well on test cases forged in a speaker-centered context. Note, for instance, that one cannot put someone in a position to know that they’ve lost the lottery on merely probabilistic grounds.⁸ And they get the right verdict in Lackey’s selfless assertion cases, since none of the norms requires any particular doxastic attitude in the speaker.

It’s a little trickier to determine whether audience-centric norms can explain the awkwardness of first-person Moorean sentences—utterances of the form, “p, but I don’t know that p” or “p, but I don’t believe that p.”⁹ But there’s at least a case to be made on their behalf. Suppose that the knowledge one would otherwise gain from testimony is defeated if the audience learns that the speaker doesn’t know or even believe what they’ve asserted.¹⁰ Then asserting, for example, “p, but I don’t believe that p” would never generate knowledge that p in one’s audience, since one would always be giving them a defeater for p along with the testimony that p.¹¹, ¹² Arguably, then, these audience-centric norms get the right verdicts in the cases that have most vexed the speaker-centric literature. And so, there’s reason to think that switching the focus of assertions from speakers to audiences is exactly what the literature needs.

Nevertheless, making this switch gets us into trouble. It predicts (I will argue) that second-person variants of Moorean sentences should systematically sound awkward in the way that first-person Moorean sentences do. But although such expressions sometimes have an air of self-defeating futility, many second-person variants of Moorean sentences sound perfectly acceptable. Taken together, these observations give us a good reason to reject audience-centric norms of assertion. One should hold onto the traditional view within the literature on the norm of assertion that the epistemic position of the speaker licenses assertion rather than that of the audience. Assertoric license comes speaker-side, not audience-side.

In Sect. 2, I introduce second-person Moorean sentences and lay out the structure of the argument. In Sect. 3, present assertions of second-person variants of Moorean propositions that don’t sound awkward in the way that first-person Moorean assertions do. But in Sect. 4, I argue that this should violate our expectations conditional on the truth of audience-centric norms. That is, if audience-centric norms of assertion were true, we should expect second-person Moorean assertions to sound awkward in precisely the way that first-person Moorean assertions do. Non-awkward second-person Moorean assertions are, therefore, evidence against audience-centric norms. I turn to objections in Sect. 5 before concluding in Sect. 6.

2 Second-person Moorean assertions

2.1 Moorean sentences

First-person Moorean sentences¹³ come in (at least) two forms:

• 1. p, but I don’t know that p.

• 2. p, but I don’t believe that p.

I assume the traditional view that knowledge entails belief.¹⁴ On this assumption, the same Williamsonian (2000) explanation can explain the awkwardness of both sentences: neither conjunction can be known by the agent who asserts (or believes) the proposition expressed by a first-person Moorean sentence.

Here’s why. Suppose I know < p, but I don’t know that p. > Then I know each conjunct. That is, I know that p. And I also know < I don’t know that p > . If I know < I don’t know that p > , then I don’t know that p (because knowledge is factive). But this can’t be right: it can’t both be true that I know that p and that I don’t know that p.

Applying the assumption that knowledge entails belief, a very similar story can be told for Moorean sentences of the form < p, but I don’t believe that p >.  It’s a conjunction I can’t know. For suppose I know < p, but I don’t believe that p > . Then I know each conjunct. First, I know that p. Since knowledge entails belief, I must also believe that p. Second, I know < I don’t believe that p > . So, I don’t believe that p (because knowledge is factive). But this can’t be right: it can’t both be true that I believe that p and that I don’t believe that p. The assumption that I know the conjunction leads to a contradiction.

Nevertheless, you, reader, can know the proposition, “p, but you”—that is, I the author—“don’t believe that p.” After all, there are many true things that I don’t believe (or know). And there is no contradiction in your knowing that p and your knowing that I don’t believe it. In short, people cannot know that Moorean sentences are true of themselves, but they can know that they are true of other people. We’re necessarily better at detecting present ignorance in others than we are in ourselves.¹⁵

Likewise, second-person Moorean sentences come in (at least) two forms:

• 3. p, but you don’t know that p.

• 4. p, but you don’t believe that p.

In a reversal from first-person Moorean sentences, these second-person variants express propositions that (if true) I can know but you can’t. The reasons that you can’t know the propositions expressed by second-person Moorean sentences exactly parallel the reasons I can’t know the propositions expressed by first-person Moorean sentences. For suppose that you know (for instance) that “p, but you [the reader] don’t believe that p.” Then you know each conjunct. First, you know that p. Since knowledge entails belief, you must also believe that p. Second, you know < you don’t believe that p > . So, you don’t believe that p (because knowledge is factive). But this can’t be right: it can’t both be true that you believe that p and that you don’t believe that p. The assumption that you know the conjunction leads to a contradiction.

Moorean propositions and their ilk show that some propositions are partial. A proposition is partial just in case it is knowable by some but not by others. Partial propositions make trouble for transmitting knowledge. Second-person Moorean sentences are determined to fall on those without ears to hear: the asserted content of a second-person Moorean sentence is not available for knowledgeable uptake by the addressee.

2.2 Framing the argument

In this paper, we are going to try to learn something about the norm of assertion by reflecting on our judgments about second-person Moorean assertions. To frame the argument, it will be useful to briefly consider how the literature on the norm of assertion has responded to first-person Moorean assertions. One of Williamson’s arguments for the knowledge norm of assertion is that it correctly predicts the awkwardness of Moorean paradoxes. We have the feeling that “something is wrong” (Williamson, 2000, p. 253) with assertions of the form “p, but I don’t know that p” or “p, but I don’t believe that p.” It’s awkward to say, for instance, “The Yankees lost, but I don’t know that the Yankees lost.” And this requires an explanation. Here’s Williamson’s: if the knowledge norm of assertion is right, then assertions of those (first-person) Moorean sort are always assertorically improper because (for the reasons given in Sect. 2.1) speakers necessarily fail to know their content. And it’s not just that they happen to be assertorically improper—assertions of that form are necessarily improper. The form of such assertions guarantees it. So, the knowledge norm of assertion rightly predicts that we should have the feeling that “something is wrong” when presented with first-person Moorean assertions.

Whether or not the knowledge norm is ultimately right, the argument from Moorean paradoxes has been seen as “one of the most important” arguments for the knowledge norm (DeRose, 2002, p. 180). It’s why opponents of the knowledge norm (e.g., Lackey, 2007) often try to show that their norm too can play a role in explaining the awkwardness of first-person Moorean sentences.¹⁶

Since the systematic awkwardness of first-person Moorean sentences is evidence for the knowledge norm, this counterfactual is true: if first-person Moorean paradoxes did not sound systematically awkward in the same sort of way, that would be evidence against the knowledge norm. After all, if the awkwardness of first-person Moorean assertions satisfies the expectations generated by the knowledge norm, the (counterfactual) non-awkwardness of first-person Moorean assertions would violate them.

In Sect. 4, I am going to argue that extant audience-centric norms of assertion entail that second-person Moorean assertions systematically violate the norm of assertion: second-person Moorean assertions never put their audiences in anything like a position to know the asserted content. And, again, it’s not just that second-person Moorean assertions happen to violate extant audience-centric norms: the form of such assertions guarantees it. In other words, these norms make the same kind of prediction about second-person Moorean assertions as the knowledge norm makes about first-person Moorean assertions. If the argument in Sect. 4 turns out to be right, then the evidential relationship between the knowledge norm of assertion and the systematic awkwardness of first-person Moorean assertions should parallel the evidential relationship between extant audience-centric norms and the systematic awkwardness of second-person Moorean assertions. That is, if we find that second-person Moorean assertions do systematically sound awkward in the way that first-person variants do, then that will be evidence for extant audience-centric norms, and if they don’t, that shall be evidence against extant audience-centric norms.

I will try to make good on the case that extant audience-centric norms predict that second-person Moorean sentences are improper in virtue of their form—and so are likely to sound awkward—in Sect. 4. But before we get there, we’ll consider examples of second-person Moorean assertions in Sect. 3. We’ll find that some second-person Moorean assertions do not share the awkwardness of their first-person counterparts.

3 Non-awkward second-person Moorean assertions

In this section, I’m going to argue that second-person Moorean assertions do not sound systematically awkward in the way that first-person Moorean assertions do. I will do that by developing a series of cases that invite that judgment.

I will also try to say something about what use it might be to assert a second-person Moorean sentence. This isn’t because I think that the “okayness” or non-awkwardness of the examples I present depends on their conversational usefulness. Rather, it’s an independently interesting question what the conversational function of second-person Moorean assertions could be given that they necessarily fail to inform their audiences of the asserted content.¹⁷

First, imagine a conversation with a climate change denier. Normally, when conversationalists disagree, it’s appropriate for disputants to share their reasons to try to resolve the dispute before moving on: we tend to seek as wide a common ground as possible. But let’s suppose that in this case you can tell that a conversation about climate change will be unproductive. Your conversation partner is showing bad epistemic faith by blatantly ignoring evidence—they are actively, not just passively, ignorant.¹⁸ After offering some statistics and identifying several scientific bodies that unequivocally affirm the reality of climate change only to be stonewalled by conspiracy theories and bad faith engagement with your arguments, it becomes clear that they aren’t going to change their mind. In fact, you know your friend well enough to know that they won’t change their mind. And, frankly, you’re tired, and aren’t feeling up to that particular Sisyphean task. After listening to a couple of minutes of your friend’s arguments, you finally assert:

• 5. Global warming is really happening, even though you don’t believe it.

Indeed, if you really know your friend well enough to know they are fully committed to denying climate change, you might assert something stronger:

• 6. Global warming is really happening, even though you’ll never believe it.

Utterances (5) and (6) seem fine to assert. They aren’t awkward in the way that first-person Moorean paradoxes are, even though there’s no way your friend can come to know that they themselves don’t believe in global warming and that it’s real at the time of utterance. Admittedly, your friend could come to know (5) in the future if they eventually changed their mind about climate change¹⁹: that is, they could come to know < Global warming is really happening, even though I didn’t believe it in the past > . But “eternal” second-person Moorean sentences such as (6) have the additional feature of being such that the addressee can never come to know it. If they ever knew it, they would have to know (and so believe) that global warming is really happening. But then they could not know the (thereby false) second conjunct, namely, < you’ll never believe that global warming is really happening > .

Often, we assert in order to transmit knowledge. That cannot be the (reasonable) goal in (5) or (6), but that does not mean the assertion is futile. What, then, have you accomplished? That answer may depend on the specific relationship between you and your friend, but one thing you may have done is to change the direction of the conversation, or perhaps even to end it. You’ve supplied information that is unshareable with your interlocutor, in the sense that you can’t both know it. Dong so blatantly may suggest that you have renounced the goal of changing your friend’s mind on climate change or of coming to a shared understanding on the topic. This may be especially helpful if you, the speaker, judge that airing the addressee’s objections would be counterproductive.

Let’s consider another example. Imagine this conversation between a conservative who watches Fox News and a liberal who believes Fox News is unreliable.

• 7. Liberal: Even Fox News says that Trump’s justification for his immigration policy greatly exaggerates the risk of terrorism.

  Conservative: Wait a minute, I thought you didn’t trust Fox News.

  Liberal: Of course Fox News is unreliable, but you don’t know that.²⁰ I’m trying to make an argument that persuades you.

Alternatively, suppose that Liberal knows Conservative’s psychology well enough to know that they will never believe (and so never know) that Fox News is unreliable. Liberal might then end the exchange with an eternal second-person Moorean assertion:

• 8. Of course Fox News is unreliable, but you’ll never know that.²¹ I’m trying to make an argument that persuades you.

Liberal is trying to give evidence that meets Conservative where they are. Instead of offering the argument that they themselves would find most convincing, they are citing evidence (testimony from Fox News) that Conservative might find convincing. Assertions typically invite challenge from one’s audience. If Liberal had simply said, “Fox News is unreliable,” then they would have invited dispute over whether or not Fox News is reliable. But by embedding that claim in a second-person Moorean sentence, Liberal effectively manages to set the record straight on Fox News while simultaneously avoiding debate about Fox News.²² In this way, Liberal redirects the conversation away from Fox News and back to immigration.

Not every second-person, Moorean-style assertion is quite so blatant. Recall Christopher Robin’s conversation with Pooh:

• 9. You’re braver than you believe (Geurs, 1997).

Or imagine that Christopher Robin knows that Pooh will always underestimate his bravery—Pooh is disposed to lack self-confidence. Christopher Robin might instead assert:

• 10. You’re braver than you’ll ever believe that you are.

Less lyrically: “You’re braver than degree X, but you don’t (or will never) believe that you’re braver than degree X.” For reasons that are by now familiar, this is something that Pooh can’t know. But nevertheless, it seems fine for Christopher Robin to say so. At the very least, it doesn’t sound awkward in the way that first-person Moorean assertions do. Christopher Robin appropriately encourages Pooh thereby.

All second-person Moorean assertions are unknowable for their addressees at the time of utterance, and—importantly—eternal second-person Moorean assertions are never knowable for their addressees. Nevertheless, I’ve developed cases in this section that show that (eternal) second-person Moorean sentences are not systematically awkward to assert. In the next section, I will argue that this is not what we should expect if any of the extant audience-centric norms of assertion is true.

4 Against audience-centric norms of assertion

Following Williamson (2000), first-person Moorean propositions have been a central data point to explain (or explain away) in the literature on the norm of assertion. Second-person variants, by comparison, have largely been ignored. As we’ve seen, however, second-person variants of Moorean propositions are theoretically interesting. Eternal second-person Moorean assertions have the intriguing twin properties of (i) necessarily failing to transmit knowledge of the asserted content to the intended audience and yet (ii) not systematically and awkwardly crashing in the way that their first-person counterparts do. (For ease of exposition, I’ll drop the “eternal” qualifier in the rest of the paper.)

I will now argue that this a mark against extant audience-centric norms of assertion. Recall the three audience-centric norms with which we began this paper:

TKR: One must ((assert p) only if one’s audience comes thereby to be in a position to know p) (García-Carpintero, 2004, p. 134; 2020, p. 270).²³

PKA: One’s assertion that p is proper only if it is fit to give a hearer knowledge that p (Pelling, 2013, p. 297).

PTW: Assert [that p] only when you could provide testimonial warrant [that p] to a potential addressee (Hinchman, 2020, p. 556).

Each, I argue, is committed to the view that second-person Moorean assertions are always improper to assert. Indeed, extant audience-centric norms predict that second-person Moorean assertions are necessarily improper merely in virtue of their form. Just as norms that predict the impropriety of first-person Moorean assertions are confirmed by the systematic awkwardness of such assertions, audience-centric norms are disconfirmed by the non-awkwardness of a wide range of second-person Moorean assertions. I address each proposed norm in turn.

4.1 TKR: transmission of knowledge rule

Suppose that one should assert that p only if one’s audience comes thereby to be in a position to know that p (TKR). What does this norm predict about the propriety of asserting second-person Moorean sentences such as “p, but you will never believe that p?” The TKR rule predicts that such assertions will always be improper. For it’s impossible for one’s intended audience to thereby come to know any proposition of that form. And if it’s impossible for an agent to know a proposition (whatever they believe), then they are not in a position to know it.²⁴

If an agent is in a position to know a proposition, then they could know it from their epistemic position. For instance, they could know it without significantly changing the evidence²⁵ upon which they could base their belief. In slogan form, the “position to know” indicates knowledge minus belief. The agent may not know what they are in a position to know, but they could know it if they believed it in the right way. But notice: even if an addressee were to believe < p, but you (that is the addressee) will never believe that p > , they would not know it. More than that, they would necessarily fail to know the conjunction, even if believed. No matter what the addressee believes or on what basis they believe, they won’t be able to know a second-person Moorean proposition. Such knowledge is not accessible to them. They are not, then, in a position to know < p, but you will never believe that p > . TKR thus predicts that second-person variants of Moorean propositions should be unassertable, merely in virtue of their form.

4.2 PKA: provision of knowledge account

Suppose that one should assert that p only if it is fit to give a hearer knowledge that p (PKA). If this norm holds, could it be assertorically proper to assert a proposition of the form < p, but you don’t believe that p > ? It seems not. For plausibly, if it’s in principle impossible for an agent to know that p, then it is not fit to give that agent knowledge that p.

But perhaps this is moving too fast: doesn’t it depend on what precisely is meant by “fitness.” Does what Pelling means by “fitness” automatically rule out second-person Moorean assertions? Here, things get tricky. Pelling defines fitness to give knowledge in terms of faultiness: an assertion is fit iff it is not faulty (Pelling, 2013, p. 303). But Pelling only gives us a definition of evidential faultiness: an assertion is evidentially faulty iff something about the “evidential basis explains why it doesn’t give knowledge in the normal way” (Pelling, 2013, p. 302). And the way in which second-person Moorean assertions are unfit to give knowledge has nothing to do with evidence. Notice, for instance, that the evidence I (the speaker) have for the proposition < p, but you don’t believe that p > could be arbitrarily strong: I (the speaker) could know it with rational certainty. But no matter how strong my evidence, you will never be able to know it.²⁶

So, if second-person Moorean propositions are faulty with respect to giving an addressee knowledge, they are faulty in some non-evidential way. Pelling only discusses ways that assertions can be evidentially faulty or unfit to give knowledge. But since second-person Moorean sentences clearly can’t give knowledge to their addressees, they seem to be faulty or unfit in some way: perhaps they are propositionally faulty in virtue of there being something about the proposition (or proposition-addressee pairing) that explains why it does not give knowledge in the normal way.

Although this explanation of the way that second-person Moorean sentences are unfit to give their addressees knowledge goes beyond Pelling’s official account, it agrees with Pelling’s overall gloss on the concept of fitness. Why introduce a notion of “fitness to give knowledge” in the first place? Because (Pelling rightly notes) an agent might assert that p very quietly, or in a language the addressee doesn’t understand, or in an epistemic context in which there is too much misleading evidence for the addressee to take the speaker’s word, and so on (Pelling, 2013, p. 296). In these sorts of situations, it is fit to give the addressee knowledge. They would be in a position to know it in better auditory, conversational, or epistemic contexts even though, contingently, the addressee won’t actually end up in a position to know the asserted content.

Intuitively, second-person Moorean propositions are radically unlike the sorts of cases that motivate Pelling’s introduction of “fitness.” It’s a necessary feature of second-person Moorean assertions—not a contingent fact about the local context—that addressees cannot come to know them. It seems, then, that such assertions should count as unfit to give knowledge.

And, indeed, elsewhere, Pelling argues that a certain proposition is fit to give knowledge because if the addressee forms a true belief in the normal way on the basis of the speaker’s testimony that the proposition is true then the addressee would also thereby acquire knowledge (Pelling, 2013, pp. 310–311). But second-person Moorean propositions necessarily fail this test. Pooh could come to truly believe, “I am braver than I believe,” not recognizing the incoherence this commits him to²⁷—he is, after all, a silly old bear—but Pooh would not thereby come to know the proposition, which is unknowable for him. That second-person Moorean sentences necessarily fail this positive test for fitness is further evidence that they are unfit to give an addressee knowledge.

Where does this leave us? We’ve been trying to evaluate whether or not it is fit to give knowledge of a second-person Moorean proposition to an addressee. I’ve suggested that the intuitive answer is no. After all, it’s conceptually impossible for the addressee to come to know it. And to borrow a test that Pelling sometimes employs, if the addressee were to truly believe it on the basis of testimony they would not thereby come to know it. Admittedly, the reason knowledge isn’t achieved in such cases doesn’t have to do with evidential deficiencies, but that (I suggest) is because evidential deficiencies aren’t the only way fitness to give knowledge can be revoked. Thus, given the PKA, second-person Moorean sentences are unfit for giving knowledge and improper to assert merely in virtue of their form.

4.3 PTW: provision of testimonial warrant

Let us now turn to Hinchman’s norm: one should assert that p only when one can provide testimonial warrant that p to a potential addressee (PTW). There are two notable ways that Hinchman’s norm differs from previous audience-centric norms: (1) the shift from knowledge to testimonial warrant, and (2) the shift from actual to potential addressee. Will either of these adjustments rescue audience-centric norms?

Hinchman shifts from knowledge to testimonial warrant, but there are a number of other epistemic properties weaker than knowledge that an audience-centric account could shift to in order to avoid the challenge from second-person Moorean assertions. I will address this class of views in Sect. 4.4. But first, I’ll say something brief about why we should be suspicious that the shift to warrant in particular will solve the problem.

Even though warrant is weaker than knowledge, it is plausible that it is conceptually linked to knowledge. Hinchman writes that “the core thought” behind his proposal is that “the normative aim of assertion is not simply to express knowledge but to give your addressee knowledge” (Hinchman, 2020, p. 579, note 12). This strongly suggests that warrant, in Hinchman’s sense, should enable addressees to acquire knowledge, at least in favorable epistemic contexts. But even if assertions of the form < p, but you don’t know that p > are warranted, in some sense, they are not warranted in any way that contributes to giving the addressee knowledge of the asserted content. Even in the best of epistemic circumstances, assertions of second-person Moorean sentences never enable addressees to know the asserted content.

Either “providing testimonial warrant that p to an addressee” entails that—in optimal epistemic circumstances—the addressee can come to know that p or else it doesn’t. Warrant of the sort that doesn’t even potentially put an addressee in a position to know doesn’t fit well with Hinchman’s explicit normative aim of giving the addressee knowledge. But if having testimonial warrant does entail at least the possibility of being in a position to know, then no addressee can have testimonial warrant for a second-person Moorean assertion. So, it seems that Hinchman’s conception of warrant in particular will either (a) be too weak to secure the core knowledge-transmitting thought behind his proposal or else (b) face the same challenge from second-person Moorean assertions as audience-centric accounts that feature knowledge.

We’ll reconsider weaker kinds of epistemic access in Sect. 4.4. For now, we turn to Hinchman’s second refinement from actual addressee to potential addressee.²⁸ In second-person Moorean assertions, although one’s intended audience cannot come to know (or be warranted in believing) the asserted content, third parties can. For instance, although Pooh cannot learn (when addressed to him) that “You are braver than you believe,” Piglet can learn the same proposition in a different guise: “Pooh is braver than he believes.” So, perhaps Christopher Robin is licensed to tell Pooh “You are braver than you believe,” not because Pooh (the actual addressee) could receive testimonial warrant but because Piglet (a potential addressee) could.

This maneuver successfully deflects the objection as (so far) developed for audience-centric accounts of assertion. But the problem quickly resurfaces. For we can imagine cases in which permissible second-person Moorean assertions rule out any potential addressee coming to learn the asserted content. Consider Cassandra, to whom Apollo gave both the gift of prophecy and the curse of never being believed when prophesying. When Cassandra foretells the destruction of Troy, she asserts without awkwardness and with apparent propriety even though, given her curse, she cannot be believed. She might assert:

• 11. Troy will fall, even though no one shall ever believe me!

No potential audience could come to know that proposition. And yet it’s not awkward in the way that first-person Moorean assertions are.²⁹

Hinchman nuances audience-centric norms in two interesting ways: shifting from knowledge to warrant and from actual addressee to potential addressee. The shift to a potential addressee is not sufficient to block the challenge from second-person Moorean assertions. And there’s textual reason to think that Hinchman’s conception of warrant is not removed enough from knowledge to neutralize the threat of second-person Moorean sentences.

This concludes our survey of extant audience-centric norms of assertion. Each norm has the consequence that certain second-person Moorean assertions are improper to assert merely in virtue of their form. In other words, they make the same prediction about second-person Moorean assertions that the knowledge norm makes about first-person Moorean assertions. If any of the extant audience-centric norms of assertion were true, we would expect, therefore, second-person Moorean assertions to sound awkward in roughly the way that first-person Moorean assertions do. But this isn’t the pattern we observed in Sect. 3. In fact, second-person Moorean assertions needn’t be awkward to assert. This observation disconfirms extant audience-centric theories.

4.4 Providing less than knowledge

This paper has now provided an argument against extant audience-centric norms of assertion: none, I have argued, can easily account for the fact that some second-person Moorean assertions are free of the awkwardness that affects their more famous first-person cousins.

All extant proposals of audience-centric norms involve knowledge, at least (as in Hinchman’s case) indirectly. And that is what makes unknowable (for the addressee) second-person Moorean sentences so tricky to handle on these accounts. This is no accident: there’s good reason to think that knowledge and assertion are, in some way, intimately related.³⁰ But I now want to suggest—somewhat more tentatively—that second-person Moorean assertions and their ilk also make trouble for yet-to-be-defended audience-centric norms of assertion. In particular, I want to suggest that audience-centric norms of assertion with this form:

General AC-Norm: Assert that p only if, by so asserting, one potentially enables one’s addressee to ϕ-ly believe that p.

where ϕ-ly believing represents some kind of epistemic access for the addressee toward a proposition (e.g. rationally believing that p) face significant difficulty. Even when ϕ is a weaker kind of epistemic access than knowledge. Obviously, this paper cannot consider every possible way of filling in ϕ, nor does this paper offer a proof that no such ϕ exists. But I shall articulate what I take to be a challenge that even non-knowledge-involving audience-centric norms of assertion must face. I suggest that the prospects for successfully answering that challenge are dim.

Here, in short, is the challenge. To be plausible, an audience-centric norm of assertion must involve potentially enabling one’s addressee to believe in a way ϕ that exemplifies a substantial enough kind of epistemic excellence. But it must also be possible, I will argue, for an addressee to ϕ-ly believe that p while knowing full well that one’s belief that p is false. And that, the argument goes, places a tight squeeze on admissible ways of ϕ-ly believing. For it’s hard to say anything very epistemically excellent about believing what one knows one must believe falsely (if one believes at all).

Let’s start with the first prong. Any plausible version of an audience-centric norm of assertion must meet the following condition:

EEE: If there is a true instance of the General AC-Norm, then proper assertions must potentially enable addressees to ϕ-ly believe the asserted content in a way that Expresses (significant enough) Epistemic Excellence.

That is a bit vague. But we can begin to see what EEE is driving at by considering some examples of norms that do not demand enough potential excellence on the part of addressees. Consider this norm:

Belief AC-Norm: Assert that p only if, by so asserting, one potentially enables one’s addressee to believe that p (in any way).

This way of filling out the General AC-Norm leaves ϕ open: any way of believing that p qualifies.

There’s a good reason no one has defended this norm. One can assert nearly anything and satisfy the Belief AC-Norm. At least if one’s audience is gullible enough. One could lie, guess, assert first-person Moorean propositions, assert what is unlikely on the speaker’s evidence to be true, and so on. Even when one asserts for no good reason whatsoever, there is (almost always) the potential for one’s addressee to believe just on the basis of one’s say-so. Believability is too low a bar for proper assertability—it licenses saying (almost) anything.

Here is a (slightly) more demanding norm:

Some Reason AC-Norm: Assert that p only if, by so asserting, one potentially enables one’s addressee to believe that p for a reason.

This norm does require some positive evaluation of the addressee’s (potential) resulting belief: the belief must be held for a reason. But this norm, too, is not demanding enough. For addressees almost always have some reason to believe the content of what is asserted: if nothing else, that they have received testimony that the asserted content is true! And so, again, this norm licenses saying almost anything to an addressee: one can lie, guess, and assert what is unlikely on the evidence. Enabling potential belief for some reason or other is also too low a bar for proper assertability.

Of course, no one has defended such weak audience-centric views. The question here is why not. The AC-theoritst’s answer: If the norm of assertion does not require addressees to believe the asserted content in a sufficiently good way (epistemically speaking), then the bar for proper assertability is set too low. The norm will license too much. The norm of assertion is supposed to play a role in explaining why we are licensed to take testimony seriously. But there is little reason for addressees to take assertions that meet such a low epistemic bar seriously at all. Thus, we return to EEE, the idea that, if an audience-centric norm is true, it must require that addressees are enabled to believe the asserted content in a sufficiently excellent sort of way:

EEE: If there is a true instance of the General AC-Norm, then proper assertions must potentially enable addressees to ϕ-ly believe the asserted content in a way that Expresses (significant enough) Epistemic Excellence.

This narrows our search somewhat. Plausible kinds of potential epistemic success that the General AC-Norm might try (among others) include truly believing, warrantedly believing, justifiedly believing, rationally believing, and reasonably believing. Indeed, the literature on norms of assertion is filled with speaker-centric norms of assertion that target these kinds of epistemic accomplishments, so this seems like a promising neighborhood for audience-centric theorists to explore.³¹

Let’s turn, now, to the second prong of the challenge for audience-centric views. Suppose some way of ϕ-ly believing p meets condition EEE. It must nevertheless be possible for an addressee to ϕ-ly believe that p while knowing full well that one’s belief that p is false.

For this argument, we need to consider some new variants of second-person Moorean sentences. In Sect. 2.1, I showed that some assertions are unknowable for addressees. It’s also true that some assertions are such that addressees cannot truly believe them. Suppose that I assert:

• 12. p, but you will never truly believe that p.

Or, perhaps more perspicuously³²:

• 13. p, but you will never have a true belief that p.

Is this something you can truly (that is, accurately) believe? No. For suppose you truly believe < p, but you will never have a true belief that p > . Then you truly believe each conjunct. So, you truly believe that p. But then it’s false that you will never have a true belief that p. So, you cannot truly believe the second conjunct (if you truly believe the first). You cannot (even potentially) truly believe both conjuncts together.

But these are just slightly modified forms of utterances that we’ve already seen are acceptable. Note that it is not assertorically awkward to assert (to the entrenched climate change skeptic):

• 14. Global warming is happening, even though you will never truly believe so.

• 15. Global warming is happening, even though you will never believe the truth about it.

So, our first observation is this: just as some assertions are not awkward in virtue of their form even though addressees cannot (even potentially) know them, some sentences are not awkward in virtue of their form even though addressees cannot (even potentially) truly believe them.

A second observation: this judgment of acceptability does not change if we stipulate that the addressee knows (and knows of themselves, de se)³³ that such assertions are not truly believable for them. So, suppose that my addressee is an entrenched, global warming skeptic who is well-versed in Moorean sentences and their ilk. It still seems fine to tell them:

• 16. Global warming is happening, even though you will never truly believe so.

And this even though it is common knowledge between us that they cannot possibly truly believe (or know) the conjunction that I assert. Such assertions may be directed at the philosopher and the layperson alike.

These two observations allow us to draw some conclusions. First, any audience-centric norm that entails true belief (i.e., that is factive) are disconfirmed by (non-awkward) second-person Moorean assertions. That’s because any way of ϕ-ly believing that entails truly believing will worryingly predict all the just-considered variants of second-person Moorean assertions (such that addressees cannot come to truly believe them) are improper merely in virtue of their form. But as we’ve just seen, they need not be awkward to assert.

Nevertheless, such assertions are not only problematic for factive, audience-centric norms. For our observations place a constraint on the kinds of epistemic success that an AC-norm can plausibly require an addressee to (potentially) have. For we’ve just seen that one can assert some proposition p even when the addressee knows that they themselves cannot truly believe p. This means that whatever epistemically excellent kind of belief an AC-norm requires addressees to (potentially) have in order to license proper assertion, it must be the kind of accomplishment that is compatible with knowing that one’s (potential) belief is false.³⁴

But this places a very tight squeeze indeed on the ways on can plausibly fill out the General AC-Norm. For it’s hard to see anything epistemically excellent about believing something that one knows one can only believe falsely. Whatever can be said positively on behalf of believing that p seems to be outweighed by the knowledge that one’s belief that p would be false.

Let’s regroup. In Sects. 3.1, 3.2 and 3.3, I argued that no extant versions of audience-centric norms could explain why (eternal) second-person Moorean assertions don’t systematically sound awkward. In this section, I’ve considered the prospects for audience-centric norms of assertion more generally that fit the following mold:

General AC-Norm: Assert that p only if, by so asserting, one potentially enables one’s addressee to ϕ-ly believe that p.

Although I don’t take myself to have definitively shown that no ϕ could possibly satisfy the formula, I do think there is a serious challenge for such norms. For to be plausible, ϕ-ly believing that p must exemplify a significant enough epistemic excellence on the part of the addressee (EEE). But it’s not necessarily awkward to tell addressees things that (just in virtue of the proposition’s form) they can’t come to truly believe. Even when addressees know that they can’t truly believe it. And that places a tight squeeze on what “ϕ-ly believing that p” could be. For it’s hard to see what could be epistemically excellent about believing what one knows will be false if one believes it.³⁵Thus, the prospects for new audience-centric norms of assertion look dim.

5 Objections

In this section, I briefly consider two objections: (1) that non-audience centric norms cannot explain why some second-person Moorean sentences do sound awkward, and (2) that second-person Moorean “assertions” may belong to some other speech-act type.

5.1 Awkward second-person Moorean assertions

I have argued second-person Moorean assertions are not systematically awkward and that audience-centric norms of assertion cannot easily explain this. But then again, there is something odd about some second-person Moorean sentences. For instance, suppose we are chatting on the phone about the weather, and I say:

• 17. It’s raining over here, but you don’t believe that it’s raining.³⁶

This is a decidedly odd thing to say. I’ve argued that audience-centric views of assertion cannot explain the apparent propriety of certain second-person Moorean assertions. But if I can’t explain the awkwardness of second-person Moorean assertions like the one above, perhaps we’ll have to call it a draw.

Fortunately, there are ready explanations in the literature for why assertions like “It’s raining over here, but you don’t know that it’s raining,” are (typically) awkward. And those explanations don’t imply that there’s anything awkward or problematic about the cases I invoke in Sect. 3.

As Sorensen notes, although some second-person Moorean sentences sound odd, they don’t necessarily sound odd in the way that first-person sentences do:

In the first person examples [of Moorean assertions], the speaker seems to be contradicting himself. In the second person examples, the speaker seems to be saying something which is self-defeating (Sorensen, 1985, p. 491).³⁷

These two different phenomenologies—the self-contradictory phenomenology and the self-defeating phenomenology—point to two different explanations of awkwardness. Moreover, there’s a ready explanation (I will soon argue) for why some second-person Moorean sentences are self-defeating and others are not: In short, self-defeating second-person Moorean sentences are those in which speakers are (or expect to be) believed and non-defeating ones are not.

When talking with my friend over the phone about the weather, it’s natural to suppose that, barring some special history of lying or joking about the weather, they will take me at my word when I tell them that it is raining in my region. I expect to be believed. When I am believed, my assertion < It’s raining here, but you don’t believe that it’s raining > has the immediate effect of making itself false. For if my friend believes < It’s raining over here, but you don’t believe that it’s raining > , then they believe < It’s raining over here > . But then my assertion is false (it’s not true that my friend doesn’t believe that it’s raining over here). And so, in contexts where my addressee will take me at my word (that is, believe what I say), asserting < It’s raining over here, but you don’t believe that it’s raining > is an example of what Hintikka (1962, pp. 90‒91) calls an “antiperformatory” assertion: asserting the proposition makes it false. Any factive norm of assertion, therefore, can rule it out.³⁸

But when speakers are not believed at their say-so, second-person Moorean assertions do not have this self-defeating character. Since the addressees do not come to believe p upon being told so, telling the addressee that p doesn’t make it false that the addressee does not believe p. This is exactly where non-awkward second-person Moorean assertions can be found. Indeed, the examples already discussed in Sect. 3 were of this sort. But here is another clear case. In The Lord of the Rings, Frodo meets a mysterious and grim-looking character named Strider who claims to be a good guide to Rivendell—a place Frodo desperately needs to go. But wary of accepting an unknown guide, Frodo demands of Strider an account of himself, to which Strider wryly responds, “[W]hy should you believe my story, if you do not trust me already? Still here it is” (Tolkien, 1984, p. 178). Strider might just as well have said,

• 18. You won’t believe what I’m about to tell you about myself (namely p), but p.

This is a second-person Moorean assertion. In context, (18) is a perfectly sensible thing to assert. And it’s sensible to assert at least in part because Strider does not expect Frodo to believe the story about himself or to be able to know that his story is true just by his say-so. And indeed, in each of our original cases, there is at least some doubt as to whether the speaker will be believed. The climate change denier won’t stop being a climate change denier at their friend’s say-so, for instance.

In Sect. 4, I argued that audience-centric norms of assertion cannot easily explain why some second-person Moorean assertions sound fine. The dialectical force of this observation against audience-centric views might be lessened if I could not explain why it is that some second-person Moorean assertions do sound awkward. But drawing on Hintikka (1962) and Sorensen (1985) has allowed us to distinguish between self-defeating second-person Moorean assertions (when speakers are believed at their say so) and non-self-defeating second-person Moorean assertions (when speakers are not so believed). The former are awkward because they make themselves false. Importantly (for the argument of this paper), this explanation leaves intact the propriety of those second-person Moorean assertions whose addressees are not prepared to take speakers at their word: such assertions are not self-defeating and do not become false upon utterance. Audience-centric norms of assertion still owe us an explanation of why such assertions are not systematically awkward.

5.2 Second-person Moorean non-assertions?

Here is a second objection. The apparent assertions of this paper aren’t assertions at all: they only look that way because they are in the indicative.³⁹ When I say, e.g., “Global warming is really happening, even though you don’t believe it,” I may be stopping the conversation or expressing my frustration that we can’t communicate about a topic I care about, but I’m not asserting. Any real assertion (the audience-centric objector insists) would involve aiming to get the audience to be positioned to know what I say thereby!

But this is unpersuasive. Even if the utterances are stopping or expressing something, they are accomplishing that task by asserting something. The reason the conversation (sometimes) stops or frustration (or another attitude) is expressed is because the audience takes the speaker to have really taken a stance on the asserted content—a stance that the addressee cannot (knowledgeably) share.

Moreover, there’s good reason to think that second-person Moorean sentences can be genuine assertions. First, we typically take first and third person versions of Moorean sentences to be assertions, so it would be odd if second-person variants were exceptional. If we didn’t take first-person Moorean sentences to be assertions, their awkwardness wouldn’t shed any light on the nature of assertion. But the awkwardness of first-person Moorean sentences is frequently interpreted as evidence for some norms of assertion and against others.

Second, uttering second-person Moorean sentences seems to enable knowledge by testimony to bystanders who overhear an assertion but are not among the audience. Suppose that Christopher Robin tells Pooh, “You’re braver than you believe” in the presence of Piglet. As noted before, Pooh cannot learn the content of the sentence, but Piglet can. That is, Piglet can learn that Pooh is braver than Pooh believes. Moreover, Piglet seems to learn this in the normal way that one acquires knowledge through (overhearing) testimony. And this is hard to explain if Christopher Robin has not genuinely asserted something with the content “Pooh is braver than Pooh believes.” If Christopher Robin were merely expressing, “Hooray for brave Pooh!” or if Christopher Robin were merely recommending that Pooh increase his estimation of his own bravery (as one could do for pragmatic reasons as well as veritistic ones), then it would be an awful coincidence that Piglet is in a position to learn, apparently by testimony, the very proposition Christopher Robin would otherwise be asserting.

6 Conclusion

In Sect. 4, I argued that if any of the extant proposals for audience-centric norms of assertion are true, then (eternal) second-person Moorean propositions are improper to assert. Given that this impropriety would be determined by the form of second-person Moorean assertions, and given the way the assertion literature has treated first-person Moorean sentences, we would expect the following to be true: if any of the extant audience-centric norms are true, then second-person Moorean assertions will sound systematically awkward. But in Sect. 3, I argued that this is not the case. Some (eternal) second-person Moorean propositions seem fine to assert. This is strong evidence against extant audience-centric norms of assertion.

Audience-centric norms like Pelling’s, García-Carpintero’s, and Hinchman’s are a helpful corrective to a literature that has too often excluded concern for the hearer, but they do so at the expense of the speaker. Assertion is (typically) an activity for pairs: the speaker and the hearer. And, indeed, it’s often true that by asserting a proposition, speakers aim to put their audiences in a position to know the asserted content.

But, on reflection, I do not think we should be so surprised if the norm of assertion allows us to assert in ways that do not line up with the typical aims that asserters have. Consider driving. If there is a typical aim of driving, it is plausibly to get people where they want to go. Certainly, we would not have developed sprawling highway systems and intricate vehicular regulations if driving were not a very useful mode of transportation. But once in place, one can pursue less typical aims while abiding by the norms of driving. One can joyride. Or take a passenger where they don’t want to go (like a child to the dentist). One can even drive aimlessly, as a pedestrian might wander. Even if the norms of assertion are principally meant to facilitate knowledge-transmitting aims, they can host a much wider variety of speaker-aims.⁴⁰

As joyriding does for driving, second-person Moorean assertions showcase appropriate, if perhaps atypical, aims that someone might have for making an assertion. As suggested in Sect. 3, they can help speakers to end a conversation or to advance an argument that meets the audience where they are, for instance. When Christopher Robin tells Pooh, “You’re braver than you (will ever) believe,” he succeeds in the aim of encouraging Pooh even though he necessarily fails to get Pooh to know the asserted content.⁴¹ Assertion may have developed as a practice for transmitting knowledge,⁴² but the practice is elastic, allowing speakers to permissibly pursue a wide range of creative purposes.

The aims of assertion are often hearer-directed, but assertoric license comes speaker-side: the speaker’s epistemic position is the grounds that enables proper assertion to the audience. Audience-accommodating norms—like Lackey’s, McKinnon’s, and mine—enable flexible transmission of information by not requiring that speakers attain the same epistemic success they hope to impart to their hearers; moreover, they do this while avoiding the false predictions that audience-centric norms encounter in second-person Moorean sentences by maintaining that it is the epistemic position of the speaker that licenses proper assertion.⁴³