2. BOB AND CAROL AND TED AND ALICE

Cases of the relevant sort are quite familiar. Here is a pair of cases of the familiar kind:

At the office, Carol asserts (1); at Alice's, Bob asserts (2) and (3):

1. (1) I know that the cupcakes don't have cinnamon in them.

2. (2) I don't know that the cupcakes don't have cinnamon in them.

3. (3) Carol doesn't know that the cupcakes don't have cinnamon in them.

Considered with respect to the conversation where it's uttered, each assertion is natural and seemingly true. A moderate will say that given Carol's epistemic position, (1) is true. Invariantism implies that (1) and (3) are contradictory, so if (1) is true, (3) is false.Footnote ⁵ Moreover, Carol and Bob have the same basis for believing that the cupcakes don't contain cinnamon, and Bob has no evidence against this proposition that Carol lacks. If their epistemic positions with respect to the cinnamon proposition are identical, purism implies that Carol is in a position to know it only if Bob is. Hence given purism, (1) makes (2) highly doubtful. But if (1) is true and (2) and (3) are false, why do (2) and (3) seem true when considered with respect to Bob's conversation?

Previous attempts to answer this question on behalf of moderate traditionalism fall into three families. Pragmatic solutions rely on the idea that assertions typically convey information beyond the contents asserted to explain our false knowledge denials.Footnote ⁶ If false knowledge denials convey relevant information in high stakes contexts, they may be appropriate even though false. One challenge for this strategy is that high stakes knowledge denials seem true, and not merely appropriate. There is no general tendency to mistake conversational propriety for truth: assertions that are hyperbolic, meiotic, metaphorical, ironic, and sarcastic can all be appropriate while clearly appearing, upon even momentary reflection, to be false. In these cases, attention to the distinction between the literal content of the speaker's assertion and what she communicates by making the assertion makes it easy to see that the former is false. Moreover, unlike standard examples of appropriate but false assertion, high stakes knowledge denials seem true even in thought. From the perspective of a high stakes situation, subjects seem not to know. It is not clear how the communicative effects of an assertion should affect how its content appears in silent contemplation.Footnote ⁷

Timothy Williamson instead offers an illusion solution. He says that when we focus on unusual possibilities of error, we tend to give them more weight in our judgments about knowledge than is warranted, thereby experiencing ‘an illusion of epistemic danger’ (Williamson Reference Williamson2005: 225–6).Footnote ⁸ But this doesn't seem like a plausible diagnosis of Bob's case. Illusory appearances are resilient even for subjects who recognize them to be misleading. Bob's attitude towards the risk that the cupcakes contain cinnamon doesn't rest on such an appearance. With additional confirmation that the unlikely cinnamon scenarios he mentions didn't occur, the epistemic appearances would change. For example, he would treat a lab test indicating that the cupcakes didn't contain cinnamon as conclusive, and so regard himself as knowing. More generally, his assessment of his epistemic situation indicates high cognitive function and epistemic responsibility, unlike a faulty judgment based on an illusory appearance. He methodically reviews his evidence and responds to Carol with a degree of caution that seems entirely appropriate given the high stakes involved.

In addition, Williamson's proposal implies that the speaker overestimates the likelihood that the subject has made an error. But this description is also inapt in Bob's case. He acknowledges the chance that he's wrong to be quite small. The high stakes lead him to take his epistemic risks seriously, but taking small risks seriously is not the same as treating them as larger than they really are. Rather, his mistake, from the perspective of the moderate traditionalist, is to judge that the tiny chance of error is incompatible with knowing.

Lost attitude solutions say that when the stakes go up, subjects tend to lose the doxastic attitude required for knowledge. Jennifer Nagel has worked out a version of this solution in brilliant detail. Nagel says that subjects in high stakes situations experience a high ‘need for closure’; i.e., they require more information before forming settled beliefs than do subjects in low stakes situations (Nagel Reference Nagel2008). She calls the force underlying this phenomenon ‘epistemic anxiety’ (Nagel Reference Nagel2011). Perhaps, experiencing heightened epistemic anxiety about the cinnamon, Bob refrains from forming a settled belief in the proposition that the cupcakes are cinnamon-free. If he lacks a settled belief, he doesn't know, and so (2) is true.Footnote ⁹

One objection to Nagel's proposal is that Bob seems capable of regarding himself as having a settled belief while still thinking that he doesn't know. He could naturally say that while he believes that there's no cinnamon, he doesn't know.Footnote ¹⁰ Moreover, the lost attitude account doesn't explain why it seems to Bob that Carol doesn't know. Nagel proposes accounting for the third-person appearance of ignorance in terms of a bias called ‘epistemic egoism’ that leads us to regard others as if they had more in common with us than they do when attributing mental states to them (Nagel Reference Nagel2010). Bob and those of us sympathetically imagining his conversation may treat Carol as if she ‘shared our heightened need for evidence’ (Nagel Reference Nagel2011: 425). This egoistic projection could result in a mistaken impression that (3) is true; Carol would appear to have adopted the attitude of knowing without having acquired the additional evidence demanded by her (projected) heightened epistemic anxiety.Footnote ¹¹ Though ingenious, the proposal is tendentious. I find no impression that Carol is in a state of heightened epistemic anxiety. Nor do I feel any inclination to treat her attitude as epistemically irresponsible: given her epistemic position and practical situation, she seems to have just the doxastic attitudes she ought to have. Of course if Nagel is right that our judgments are distorted by a cognitive bias we should be alert to the possibility that our introspective assessment of the basis for those judgments is inaccurate. Still, it is desirable to have an explanation for (3)’s apparent truth that does not rest on the claim that we treat Carol as being either in a state of heightened epistemic anxiety or epistemically irresponsible.

Moreover even granting Nagel's egoistic projection explanation for the apparent truth of (3), the problem of false denials persists. Bob also seems not to be in a position to know. From the perspective of his conversation, it seems that even if he did have the attitude required for knowledge, he still wouldn't know. Indeed, in his conversation with Alice an assertion of (2a), though a bit clunkier, would be just as apparently true as his assertion of (2):

1. (2a) I'm not in a position to know that the cupcakes don't have cinnamon in them.

If Bob lacks the attitude of knowledge, perhaps this is because it appears to him that his epistemic position doesn't warrant it. The moderate traditionalist must treat this appearance as misleading. And in response to Alice's last question Bob could assert:

1. (3a) Carol isn't in a position to know that the cupcakes don't have cinnamon in them.

Given that (1) is true, (3a) is false. Thus we have another false appearance to account for: that, from the perspective of the high stakes conversation, neither Bob nor Carol is in a position to know.

3. SIX DESIDERATA FOR A SOLUTION

The challenges faced by the three solutions just discussed reflect six aspects of the high stakes appearance of ignorance that an adequate moderate traditionalist solution should account for:

A fully satisfactory solution to the problem of false denials should explain each of these aspects. The pragmatic solution struggles with Persistence and Thought; the illusion solution leaves Responsibility unexplained and – on Williamson's formulation, at least – is incompatible with Accuracy; the lost attitude solution predicts neither Projection nor Position to Know. In what follows, I'll lay out a solution that predicts and explains all six aspects of the high stakes appearance of ignorance. But first, a detour through a powerful argument against moderate traditionalism.

4. UNASSERTABILITY AND IGNORANCE

The apparent truth of (2) and (3) can be bolstered by an argument.Footnote ¹² It is plausible that a speaker who knows that p is in an epistemic position sufficient for making an epistemically proper assertion that p. The sufficiency principle can be defended by appeal to the proposal that the constitutive norm of assertion is that one should assert only what one knows; if there are no other rules specific to assertion, anyone who knows that p is in an epistemic position to assert that p.Footnote ¹³ It can also be defended by appeal to any proposal about the constitutive norm of assertion that specifies a condition satisfied by anyone who knows.Footnote ¹⁴ If the constitutive rule of assertion is that one must satisfy C in order to assert that p, and C is a condition satisfied by anyone who knows that p, then anyone who knows that p is in a position to assert that p. Finally, it can be defended by appeal to our conversational practices. We recommend speakers as sources of assertions on the basis of our presumption that they are in a position to know and criticize speakers whose assertions would have been useful when we find out that they were in a position to know. Both practices suggest that we tacitly accept the sufficiency principle.Footnote ¹⁵

Yet in his conversation with Alice, Bob would clearly do something epistemically improper by asserting that the cupcakes are cinnamon-free. Having acknowledged the slight but real chance that he's mistaken, and fully cognizant of the stakes involved, he could not then properly assert to Alice that the cupcakes have no cinnamon in them. If being in a position to know suffices for being in a position to assert, then we can infer ignorance from unassertability:

1. (U1) In his conversation with Alice, Bob is not in an epistemic position to assert that the cupcakes are cinnamon-free.

2. (U2) If a speaker is in a position to know that p, then he or she is in an epistemic position to assert that p.

3. (U3) So, in his conversation with Alice, Bob isn't in a position to know that the cupcakes are cinnamon-free.

Since Carol and Bob are in the same epistemic position with respect to the cinnamon proposition, (U3) together with traditionalism implies that she is not in a position to know either.

How should the moderate traditionalist respond to this argument? It is not appealing to say that Bob would do nothing epistemically wrong by asserting the cinnamon proposition in his conversation with Alice. It is one thing to claim that Bob's high stakes knowledge denials are, contrary to appearances, false. It's another to say that Bob's assertion to Alice that the cupcakes are cinnamon-free would be epistemically impeccable. We may be mistaken about whether a subject knows without being mistaken about whether a particular speech act would be proper. If moderate traditionalism indicted our capacity to assess the propriety of assertions in high stakes situations, it would incur heavier theoretical costs than just the problem of the misleading appearance of ignorance. The more obvious line of resistance is to reject (U2). Some traditionalists do argue that knowledge is insufficient for epistemically proper assertion.Footnote ¹⁶ But is there another way for the moderate traditionalist to resist the unassertability argument? In the next section I'll offer a way to endorse each premise in a form that renders the argument invalid. This path of resistance will lead, in the following section, to a new solution to the problem of false denials.

5. HIGH STAKES UNASSERTABILITY

In doing one thing, you often do others as well. By practicing a piece on the piano you commit it to memory; by cycling down the path you commute to work; by harrumphing you express your irritation.Footnote ¹⁷ Moreover, what you do by doing something depends upon the circumstances in which you do it. In different circumstances, you would practice the piece without committing it to memory (because of squealing children tumbling through the living room); cycle down the path without commuting (because you have the day off); or harrumph without expressing your irritation (because you are acting in a play). The circumstances in which you act determine, in large part, all of what you thereby do. As Arthur Danto says, ‘We must read the identity of the action from the context of its execution’ (Danto Reference Danto1973: ix).

Assertion is no exception: by asserting you often do other things as well, though what else you do depends upon your circumstances. You can greet a friend by asserting that you're happy to see her, dismiss an objection by asserting that it rests on a claim that's been refuted, insult your host by asserting that his cooking is incompetent, or commit heresy by asserting that God is not triune.Footnote ¹⁸ In different circumstances the same assertions may have constituted different doings. In culinary school an assertion that someone's cooking is incompetent may constitute constructive criticism, not an insult; had early church councils had different outcomes, an assertion that God is not triune might have constituted an expression of Christian orthodoxy, not a commission of heresy; and so on.

With these observations in mind, we can distinguish two conceptions of an epistemically proper assertion. I'll say that an assertion is narrowly epistemically proper just in case it is epistemically proper qua assertion. An assertion's narrow propriety concerns the question: ignoring whatever else the speaker would do by asserting that p, is she in an epistemic position to assert that p? The concept of narrow propriety may be elaborated with the idea that there is an epistemic rule specific to assertion. On this way of conceiving of narrow propriety, an assertion is narrowly proper just in case it violates no assertion-specific epistemic rule. For example, if the Knowledge Rule is the only epistemic rule specific to assertion, then an assertion is narrowly epistemically proper if and only if it is known. Weaker rules that specify conditions satisfied by those who know that p will also imply that a speaker who knows that p is in a position to make a narrowly epistemically proper assertion that p.

By contrast an assertion is broadly epistemically proper just in case, by asserting, the speaker does nothing that she's not in an epistemic position to do. Our practice of assessing action in terms of epistemic propriety is hardly limited to assertion. You shouldn't venture out on to the ice without good evidence that it's thick enough to hold your weight; you shouldn't take a cake out of the oven unless you have good evidence that it's done.Footnote ¹⁹ If by asserting you do something else that you're not in an epistemic position to do, your assertion will exhibit an epistemic impropriety not traceable to any epistemic rules specific to assertion, or the propriety of your assertion qua assertion. An assertion that is broadly proper will also be narrowly proper, but since one can do something epistemically improper by asserting without doing anything improper qua asserting, an assertion that is narrowly proper may not be broadly proper.

The principle that being in a position to know p is sufficient for being in a position to assert p can thus be understood in two ways. On the narrow construal, the claim is that a speaker in a position to know p is in a position to make a narrowly epistemically proper assertion that p. On the broad construal, the claim is that a speaker who knows that p is in a position to assert that p without doing anything she's not in an epistemic position to do.

The broad sufficiency claim is not plausible. The sufficiency principle draws support from popular views about the constitutive norm of assertion, as well as conversational practices that indicate that ordinary speakers presuppose that knowing puts them in a position to assert. But while these sources support the idea that known assertions are epistemically proper qua assertions, they give us no reason to think that any act whatsoever that one performs by asserting is epistemically proper so long as the speaker knows the content she asserts. Earlier in this section I mentioned four things one may do by asserting: greet a friend, dismiss an objection, insult a cook, commit heresy. That is a very small sample. Given the infinite variety of possible conversational circumstances, it seems there is a limitless number of things one may do by asserting. It would be surprising to find that for any of the many φs one may apparently perform by asserting, it is in fact possible to φ by asserting that p only if knowing that p is sufficient for epistemically proper φ-ing.

Three families of cases furnish natural counterexamples to the broad sufficiency claim. First, by asserting that p, one may conversationally implicate information one is not in a position to communicate. For example, by asserting that there is a gas station around the corner, a speaker may implicate that gas is available around the corner. If the speaker has good reason to think that the gas station is long abandoned, she will have done something epistemically improper by asserting that there is a gas station around the corner, even if she knows that there is a gas station around the corner. Of course the speaker could cancel her implicature, and so assert without doing anything epistemically improper; the point is that if she allows the implicature to stand, by asserting she will have done something epistemically improper despite knowing the content she asserts.

Second, by asserting that p one may express an attitude towards p one ought not, epistemically, to express. For example, in some conversations an unusually demanding epistemic standard may become salient. In such a conversation a flat-out assertion that p is liable also to constitute an expression of an especially high degree of certainty that p. If the speaker knows that p but is not in an epistemic position that warrants especially high certainty that p, she will do something epistemically improper by asserting that p.Footnote ²⁰

Third, assertions may take on unusual significance by the adoption of special conventions. For example, suppose that you and your friend are poker cheats who have a code for communicating surreptitiously to each other during the course of play. One element of the code is this: whenever one of you wants the other one to fold, you're to make an observation about the temperature in the room. Of course your success as cheaters requires that you'll instruct your partner to fold only if you have good reason for doing so. During the course of the game, you notice that it's getting very warm in the room. If you're not in an epistemic position to instruct your partner to fold – you've been negligent in observing the other players' hands – you'd do something epistemically improper by asserting that it's warm in the room, despite knowing that it's warm in the room.

In each type of case, there is some p such that the speaker knows that p but would do something broadly epistemically improper by asserting that p. They illustrate that being in a position to know that p does not ensure that one's epistemic position with respect to p is strong enough to make a broadly epistemically proper assertion. But they provide no evidence against the narrow sufficiency claim. In each case, the epistemic impropriety of the assertion derives from the epistemic impropriety of something else the speaker would do by asserting. So provided the sufficiency principle is understood as a claim about narrow propriety, its plausibility is not undermined by such cases:

1. (U2N) If a speaker is in a position to know that p, then she is in a position to make a narrowly epistemically proper assertion that p.

(U2N) implies that a known assertion is narrowly epistemically proper, but not that it is proper tout court. (U1) implies that Bob would do something epistemically improper by asserting that the cupcakes are cinnamon-free in his conversation with Alice. But unless his assertion would be narrowly improper, (U2N) will not license the conclusion that he is not in a position to know that the cupcakes are cinnamon free. If we have a way to diagnose the impropriety of Bob's assertion in terms of what else he would do by asserting, we have a way to disarm the unassertability argument. Is such a diagnosis available?

6. GUARANTEEING AND OTHER EPISTEMICALLY DEMANDING CONVERSATIONAL ACTS

John Turri argues that in high stakes contexts, speakers who make assertoric utterances thereby guarantee (Turri Reference Turri2010). To properly guarantee that p, you need to be in a stronger epistemic position with respect to p than you do to properly assert that p; on Turri's favored view, proper guaranteeing requires knowing that you know. Suppose that Bob's assertion would constitute a guarantee that the cupcakes are cinnamon-free. If he doesn't know that he knows this, then by asserting he would do something epistemically improper, even if he knows and hence is in a position to make a narrowly proper assertion. I'm not convinced that Turri's `knowing-that-you-know' account of proper guaranteeing is satisfactory, so I won't rely on it further.Footnote ²¹ The key idea is that to properly guarantee that p requires a stronger epistemic position with respect to p than does properly asserting that p; in particular, that it requires more than being in a position to know that p, and something that Bob lacks.Footnote ²²

Turri's proposal gives us a way to treat Bob's assertion as merely broadly improper: though he knows that p, in his conversation with Alice his assertion would constitute a guarantee, which he is not in an epistemic position to make. If so, we have a rationale for treating (U1) as a claim about broad propriety:

1. (U1B) In his conversation with Alice, Bob is not in an epistemic position to make a broadly proper assertion that the cupcakes are cinnamon-free.

Since broad impropriety does not imply narrow impropriety, (U1B) and (U2N) do not together imply (U3). This treatment of the premises gives us a way to disarm the unassertability argument without rejecting the sufficiency principle.

One might object that if Bob's assertion would be improper only in virtue of constituting a guarantee, he should be able to assert properly by disavowing an intention to guarantee: by saying, for example, ‘The cupcakes don't have any cinnamon, though I don't guarantee it.’ But such a clarification would not make the assertion epistemically proper. Since the assertion's epistemic impropriety would survive an explicit disavowal of the speaker's intention to guarantee, it must attach to the assertion qua assertion; i.e., it must be narrow impropriety.

However, the objection makes two questionable assumptions. The first is that to assert while disavowing an intention to guarantee is sufficient to assert without guaranteeing. When circumstances are such that by asserting one would also φ, intending not to φ is not generally sufficient for asserting without φ-ing. Indeed, whether an assertion constitutes a φ-ing can lie quite outside the speaker's control. The assertions of a witness who has taken an oath to tell the truth constitute sworn testimony whether or not she intends them to. She cannot refrain from offering sworn testimony by mentally crossing her fingers. Perhaps guaranteeing is like this: in high stakes conversations one may be unable to prevent one's assertoric utterance from constituting a guarantee, even if one intends it as a (mere) assertion. If intending to assert without guaranteeing is insufficient to prevent Bob's assertion from constituting a guarantee, then disavowing an intention to guarantee won't restore his assertion to epistemic propriety.

The second assumption that the objection makes is that the impropriety of the utterance would be epistemic in nature. This may be doubted. When a guarantee is needed or expected, a mere assertion is liable to be pointless or misleading. Thus someone who asserts while explicitly disavowing the intention to guarantee in such a circumstance is apt to express a stubbornly uncooperative attitude, or else to signal confusion about the conversational demands. If, in Bob's conversation, a guarantee is needed or expected, an assertion of ‘The cupcakes don't have any cinnamon, though I don't guarantee it’ would be pragmatically defective. This defect would enable us to explain the assertion's impropriety without reference to any epistemic impropriety. If the impropriety of the conjunction can be explained in these more broadly pragmatic terms, it does not provide evidence that Bob's assertion that the cupcakes don't have any cinnamon is narrowly improper.

Turri's suggestion about guaranteeing is not the only way to diagnose the impropriety of Bob's assertion as being merely broad. There are other candidates for the epistemically improper action Bob's assertion would constitute in his conversation with Alice. For example, by asserting he may also:

• • Express an attitude of certainty towards the cinnamon proposition.

• • Promise Alice that Ted won't have an adverse reaction to the cupcakes.

• • Convince or persuade Alice to give Ted a cupcake.Footnote ²³

Recognizing that there is a tiny but conversationally relevant chance that the cupcakes have cinnamon in them, Bob is not in a position to express certainty. Similarly, he's not in a position to promise Alice that Ted won't have an adverse reaction to eating them. And given the risks, his epistemic position doesn't justify him in convincing Alice to give Ted a cupcake. Independently of Turri's proposal about guaranteeing, then, the moderate traditionalist who wishes to respect the sufficiency principle has numerous paths to resist the argument from unassertability to ignorance.

7. UNASSERTABILITY AND THE APPEARANCE OF IGNORANCE

Now we return to the appearance of ignorance. The distinction between narrow and broad propriety opens up a new explanation for the appearance of ignorance in cases like Bob's. According to the unassertability solution, Bob recognizes that his epistemic position with respect to the cinnamon proposition is insufficient for properly asserting it. He also has the sufficiency principle among his background presuppositions. His recognition of the epistemic impropriety of the assertion, together with his acceptance of the sufficiency principle, generates the impression of ignorance. If the impropriety would be merely broad, the appearance may be inaccurate. When a subject's epistemic position with respect to p is sufficient for narrow but not broad assertability, speakers are apt to experience a misleading impression of ignorance.

Not every broadly improper but narrowly proper assertion will appear unknown, even to speakers who are sensitive to its impropriety and presuppose the sufficiency principle. If an assertion that p would be broadly improper because it would generate an unwarranted implicature that q, we should not expect the speaker to judge that she doesn't know that p. The key factor in Bob's case is that the broad impropriety of asserting p results from a weakness in his epistemic position with respect to p. The unassertability solution says that when a weakness in the speaker's epistemic position with respect to p would make her assertion that p improper and she presupposes that being in a position to know is sufficient for being in a position to assert, she will tend to judge that she is not in a position to know that p. When, in fact, she is in a position to know that p – and hence her assertion would be merely broadly improper – this judgment will be inaccurate.

The solution credits Bob with awareness that:

1. (*) Due to a weakness in my epistemic position with respect to p, I should not assert that p.

As we have seen (*) can be made true in two different ways:

1. (*N) Due to a weakness in my epistemic position with respect to p, asserting that p would be narrowly epistemically improper.

2. (*B) Due to a weakness in my epistemic position with respect to p, asserting that p would be (merely) broadly epistemically improper.

In Bob's conversation, (*) is true in virtue of (*B), and not in virtue of (*N). The capacity to differentiate cases where (*B) is true from cases where (*N) is true presupposes a grasp of a philosophically sophisticated distinction that we can hardly expect ordinary speakers to draw, especially in the hurly-burly of ordinary conversation. Moreover, the distinction is of little practical significance. Whether (*) is true because of (*B) or (*N), the action-guiding upshot is the same: the speaker ought not to assert p without first improving her epistemic position with respect to p. For ordinary speakers to avoid epistemically improper assertions, sensitivity to impropriety tout court is enough.

Not only is there no practical significance in distinguishing between cases where (*N) is true and cases where (*B) is true, the distinction will remain elusive even to speakers who are sophisticated enough to draw it. When asserting that p would be broadly improper, asserting that p is known would be even more emphatically improper. So in circumstances where a speaker cannot assert that p because (for example) she is not in an epistemic position to guarantee that p, she will also be unable to assert that she knows or is in a position to know p. Given the sufficiency principle, this means that when a speaker is in a position to make a narrowly proper (because known) assertion, but not a broadly proper one (due to a weakness in her epistemic position with respect to p), she will be in no position to assert that this is the case. For such a speaker to provide an accurate description of her position vis-à-vis p's assertability would require making an assertion that is, in the circumstances, forbidden on epistemic grounds.

Even though the distinction between (*B) and (*N) is elusive and practically insignificant, blindness to it will, together with the sufficiency principle, tend to produce misleading impressions of ignorance. Bob recognizes that his epistemic position with respect to the cinnamon proposition makes it improper to assert that proposition in his conversation with Alice, and accepts that being in a position to know is sufficient for being in a position to assert. Insensitivity to the distinction between (*B) and (*N) thus leads to the impression that the cinnamon proposition is unknown.Footnote ²⁴

8. THE UNASSERTABILITY SOLUTION AND THE DESIDERATA

The unassertability solution predicts that a high stakes speaker who recognizes that she is not in a position to assert that p will tend to judge that she is not in a position to know that p: the appearance of ignorance is grounded in a recognition that her epistemic position is insufficient for assertability, which leads, via the sufficiency principle, to the judgment that she is not in a position to know (Position to Know). We should expect that subjects who make such a judgment will regard the literal contents of their knowledge denials as true (Semantics), and that these contents will appear true even in thought (Thought). Since the judgment stems from an accurate appraisal of the speaker's own epistemic position – in particular, that her position is insufficient for proper assertion – it is one we should expect epistemically responsible speakers to make (Responsibility), and to involve no overestimation of her epistemic risks (Accuracy).

Finally, the solution predicts that subjects in the same epistemic position as the speaker will also appear not to be in a position to know (Projection). A high stakes subject who succumbs to the error posited by the unassertability solution thinks that her epistemic position is insufficient for knowing. Whatever weakness in her own epistemic position presents an apparent block to knowing will also appear to prevent another subject in the same epistemic position from knowing. So we should expect her to treat any subject whose epistemic position she takes to be equivalent to her own as also not being in a position to know. The unassertability solution thus predicts that first-person and third-person judgments about knowledge and ignorance will sway together.

9. THE SOLITARY APPEARANCE OF IGNORANCE

Like most widely known cases mustered as evidence against traditionalism, those of Carol and Bob center on conversations, and so provide the raw materials for the unassertability solution to work with. But what about situations where assertion is not an issue?Footnote ²⁵ Suppose that Bob doesn't learn of Ted's cinnamon allergy from Alice, but discovers in solitude that he himself has the dangerous condition. If, considering his own epistemic situation, he judges that he is not in a position to know the cinnamon proposition, then he has made an error by the moderate traditionalist's lights. And yet he is not contemplating an assertion of that proposition. A high stakes error that is not rooted in a judgment about the epistemic propriety of an assertion isn't explained by the unassertability solution. What should we say about such cases?

First, it is not altogether clear that a subject privately contemplating the question of p is not considering an assertion that p. Solitary deliberation about p is most readily imagined as an internal dialogue concerning p. It is natural to think that an internal dialogue that issues in a judgment that p ends in a private assertion that p. So even subjects merely wondering whether p may be contemplating an assertion that p; i.e., an assertion of p to themselves. If judgment is a species of assertion, then we could apply the unassertability account to cases not involving conversation. However, even granting the thesis that solitary judgment is a species of assertion, the unassertability account could only handle such cases if a judgment that p would also constitute an act that the subject's epistemic position with respect to p does not warrant.

Still, the unassertability solution illuminates a path for explaining the solitary appearance of ignorance. The solution implements a more general strategy. The general strategy starts by identifying a principle on which being in a position to know that p is sufficient for epistemically proper φ-ing. In a case where there is no practical difference between φ-ing and ψ-ing, but to ψ would be improper due to a weakness in one's epistemic position with respect to p, acceptance of the principle can generate the misleading appearance of ignorance. A subject who is sensitive to the fact that φ-ing would be epistemically improper due to a weakness in her epistemic position and fails to distinguish between φ-ing and ψ-ing can be led by the principle to judge that she isn't in a position to know p. The unassertability account results when φ-ing is asserting that p and ψ-ing is some other conversational act that would be performed by an assertion that p. But the strategy can be implemented in other ways. Here is one possibility:

Prescinding from the characterization of judgment as a private form of assertion, it is plausible that being in a position to know that p is sufficient for epistemically proper judging that p: if you're in a position to know that p, you're in a position to judge that p. In some circumstances, though, there may be no practical difference between a judgment that p and a distinct cognitive act one is not in a position to perform. And in particular, an epistemically proper judgment that p may not be practically distinct from a cognitive act one's epistemic position with respect to p does not warrant. For example, when I am knowingly presented with a choice that will result in my death if p is false, there may be no practical difference between judging that p and resolving to risk my life on my being correct about p. But I can know that p without its being epistemically proper for me to risk my life on the question of whether p.Footnote ²⁶ Moreover, when I am not in an epistemic position to risk my life on p, that's partly because of a weakness in my epistemic position with respect to p: properly staking one's life on p requires being in a stronger epistemic position with respect to p than most of us are with respect to many propositions we are in a position to know. A subject who recognizes that by judging that p she would stake her life on p, and that her epistemic position with respect to p makes such a risk epistemically improper, will, if she treats being in a position to know as sufficient for epistemically proper judging, be susceptible to the misleading appearance of ignorance in a way closely analogous to that described by the unassertability account. It is not difficult to fit the solitary version of Bob's high-stakes impression of ignorance into this suggestion.