1. Introduction
Testimony is how beliefs get around. In paradigmatic cases of testimony, S begins with a belief that p. S asserts that p. And a hearer, H, comes to belief that p on the basis of S’s say-so. Exactly when it’s appropriate for a speaker to assert that p and under what conditions a hearer is entitled to believe p on the speaker’s say-so is contentious. But all parties agree that assertion is a paradigmatic way that beliefs get from one person to another − a way for beliefs to be transmitted through testimony.
Questions − or the speech act that I shall sometimes call queries or askings − are usually conceived differently. In paradigmatic cases of asking a question, I ask you a question not primarily to prompt you to form the question that I have but to get you to give me the answer that you possess.
Indeed, several philosophers have thought that it’s inappropriate to ask a question unless you know that the speaker already knows the answer (and can thereby inform you). Most recently, Haziza (Reference Haziza2023) has defended this norm:
Addressee-Knowledge Norm (AKN): Ask addressee A the question Q only if A knows Q. (Haziza Reference Haziza2023: 113)
At least, Haziza thinks AKN applies to what he calls inquiry questioning (Haziza Reference Haziza2023: 114), in which the speaker is inquiring into − and trying to settle − the question they ask. Haziza has offered the most rigorous defense of this norm, but he is not alone. Consider the following comparable views:
-
• “A plausible norm for questioning is that one should ask someone a question only if she knows the answer” (Schaffer Reference Schaffer2006: 98).
-
• “The questioner should have reason to believe that the questionee knows the answer” (Sorensen Reference Sorensen1981: 159).
-
• “The question ‘Is Mr. Smith at home?’ can be expressed by three sentences in the indicative mood,” one of which is “I believe that you know whether Mr. Smith is at home” (Jeffreys Reference Jeffreys1948: 378).
There are several reasons for holding such a norm. But at a high altitude, one reason to think this is that it gives the speech act of querying a satisfyingly complementary role to that of assertion. When we assert, we (typically) say what we know.Footnote 1 Assertion is a way to get knowledge from the speaker to the audience. The AKN for asking flips these roles for asking. When we ask a question, we do so because the audience can give knowledge to the speaker through an assertion that answers the question.
Understood negatively, my thesis is that this attractive picture of the speech act of questioning is mistaken. We routinely ask questions without knowing (or even reasonably believing, etc.) that the speaker knows the answer.
More positively, I argue that the reason the attractive picture above is mistaken is that asking is not merely the complement of asserting. When we assert, we put forward a proposition. When we query, we raise a question. Granted, we often raise a question in order to get an answer from our interlocutor. But we also do it sometimes to transmit the question. Just as beliefs get around through testimony, interrogative attitudes (IAs) get around through asking.
Just as we need rules governing testimonial transmission (i.e., only assert what is true), so we need rules governing interrogative transmission. Some testifiers are trustworthy informants − they (reliably enough) abide by the rules that govern testimony. We should also want to be trustworthy interrogators. But what does being a trustworthy interrogator require?
2. Against the addressee knowledge norm
In this section, I argue against the AKN. I do this, in part, because AKN is an interesting thesis in its own right that deserves consideration. But we also start here because the cases in which AKN fails are those in which it is clearest that something like interrogative transmission is going on. AKN is a demanding principle, but Haziza (Reference Haziza2023) has put together an impressive compilation of linguistic data in its favor.
2.1. Data for AKN
Haziza (Reference Haziza2023: 8) invites us to imagine a conversation between Ann, who is looking for her keys, and her coworker Bob, whom Ann has no special reason to believe would know where her keys are:
Ann: Where are my keys?
Bob: Why would I know where they are?
Ann: Do you know where my keys are?
Bob: ??Why would I know where they are?
Haziza’s judgment is that Ann’s second question is preferable. Given that she has no idea whether Bob knows where her keys are, it is inappropriate for Ann to ask him where her keys are. It is much better for her to ask, “Do you know where my keys are?” This is evidenced by the fact that Bob seems to have a valid complaint in response to the first question (“Why would I know that!”) that is not available to him in the same way in response to the second question.
Why is this evidence for AKN? Given that Ann has no idea whether Bob knows where her keys are, she doesn’t know that Bob knows the answer to “Where are my keys?” But presumably, she does know that Bob knows the answer to “Do you know where my keys are?” Thus, AKN predicts that her second question is better.
I have two objections to this argument, the first technical and the second substantive. The technical puzzle for this argument is that, given that knowledge is anti-luminous,Footnote 2 it’s not clear that Ann (always) knows whether Bob knows where her keys are. And so it’s not clear AKN actually explains why Ann’s second (indirect) question seems better. Suppose it’s a borderline case whether Bob knows where Ann’s keys are. Perhaps he vaguely remembers seeing them sitting on the reception desk, but his memory is a bit hazy, and he’s not sure that he’s right. (If this doesn’t seem like a case of borderline knowledge to you, fill in the details of the case so that it does.) In this case, Bob might not know whether he knows where Ann’s keys are. Or if you like, imagine Bob has been reading some skeptics lately, and he’s unsure whether anyone knows anything (including whether he knows where Ann’s keys are). In those cases, Ann violates AKN by asking Bob the indirect question whether he knows where her keys are: Bob, under the influence of the skeptics, doesn’t know whether he knows. But it’s doubtful that Ann violates a genuine norm by asking Bob if he knows where her keys are just because he’s in a borderline case of knowledge, or because he’s been brushing up on the Pyrrhonians.Footnote 3
The deeper worry is that our tendency to ask indirect questions is a general phenomenon: we don’t only do it when we suspect our addressee doesn’t know the answer. Just as it’d be suspicious to cite a particular policy of the Scottish government as an explanation for an economic downturn affecting not just Scotland but the whole of Europe, so there’s a worry that AKN’s explanation is too narrowly tailored given the breadth of the pattern of asking indirect questions.
Brown and Levinson (Reference Brown and Levinson1987) treat indirect questions in their seminal work on politeness. On their view, there are both positive and negative strategies for helping each other preserve face in the midst of face-threatening acts (FTAs). One important negative strategy is indirectness. A speaker is often indirect “when faced with opposing tensions: the desire to give [the hearer] an ‘out’ by being indirect, and the desire to go on the record” (Brown and Levinson Reference Brown and Levinson1987: 132).
Brown and Levinson (Reference Brown and Levinson1987: 68) list displaying “ignorance of something S is expected to know” as one way to lose face. When we ask someone a question, there is social pressure on them to respond. If they know the answer to the question, then (other things being equal) they can display their knowledge. If they don’t, they must confess their ignorance on a topic directed to them (or deviate from social expectations in other ways, e.g., by suddenly changing the topic). Accordingly, when we ask someone a question that we don’t know that they know the answer to, we generally risk imposing the loss of face on our interlocutor. Of course, there are other ways to ask questions in ways that may cause our interlocutors to lose face or suffer other harms − Berstler (Reference Berstler2023) explores some of the dangers of asking intrusive questions, for instance. Asking questions that may force our interlocutors to display ignorance is thus just one way among many that we can ask a socially risky question.
Brown and Levinson argue that indirectness is an important strategy for saving face, often giving interlocutors “a line of escape” (Brown and Levinson Reference Brown and Levinson1987: 134) to maintain face. For example, suppose I would like to ask you to drive me to the airport, but our friendship isn’t such that I am sure this is something you are happy to do. I could ask:
-
(1) Will you drive me to the airport?
But if you are not happy to do so, this puts you in the uncomfortable position of either being imposed upon (by agreeing to drive me to the airport even though you don’t want to) or, embarrassingly for both of us, revealing that helping me achieve my goal of getting to the airport is not a priority for you.
To mitigate this risk, I may engage in what Brown and Levinson call “redressive” face-preserving behavior. Suppose instead of (1), I ask:
-
(2) Are you free to drive me to the airport?
This question gives you an out. If you don’t want to drive me to the airport, you can answer that you are not free (where perhaps the addressee is given some latitude in what it takes to count as being “free”). You can answer “no” without either of us losing face. At some level, I’m not ultimately interested in whether you are free to drive me to the airport; I’m interested in whether you will drive me to the airport. Even so, we often prefer indirect, face-preserving questions over the direct questions we are ultimately animated by.
Indirect questions that begin with “do you know” similarly give addressees a face-preserving way out. When Ann asks Bob where her keys are, if Bob doesn’t know, he won’t be able to give an answer to the question he’s asked. But it’s very likelyFootnote 4 that Bob can provide an answer to whether he knows where Ann’s keys are.
It’s true that, assuming Bob doesn’t know where Ann’s keys are, he will manifest ignorance either way. If Ann asks the direct question, he will display ignorance by replying “I don’t know” in response to the question. If Ann asks the indirect question, his answer will be that he does not know where her keys are. But it’s only in the first case that Bob displays ignorance on a question where a response from him is conversationally expected. When responding to the indirect version of Ann’s question, Bob is able to knowledgeably answer every question put to him.
In short, I’ve suggested that there’s an alternative explanation for Haziza’s data − an explanation that doesn’t require AKN. Asking someone a question when they might not know the answer can be an FTA. It is thus for reasons of politeness (rather than a seemingly epistemically inflected norm like AKN) that we have a preference for indirect rather than direct questions when, for all we know, our addressees are ignorant.
Let’s look at one more piece of linguistic data Haziza leverages in support of AKN. Haziza identifies certain constructions as “Moore-paradoxical” askings.Footnote 5 Consider the following (lightly adapted) cases from Haziza (Reference Haziza2023: 9, 11):
Ann: Do you know when the meeting starts?
Bob: No.
Ann: #When does it start?
Ann: #You don’t know whether Bob is coming to the party, but is he coming to the party?
AKN elegantly explains what’s wrong with these examples; this data shows, yet again, why AKN should be taken seriously.
I tried to undercut Haziza’s argument from indirect questions by offering an alternative explanation of the same data. That explanation carries over to Moore-paradoxical askings. If it risks making your addressee lose face to ask them a question they may not be able to answer, it’s even less polite to ask such a question knowing that they don’t know the answer.
Somewhat more tentatively, I note that while Haziza’s examples of Moore-paradoxical askings do seem problematic (and rude) to me, I’m not sure they seem paradoxical in the way that Moorean paradoxes do. “You don’t know wh-Q” doesn’t obviously take back the question Q − even if it denies the addressee the ability to respond − in the way that “I don’t know p” seems to take back an assertion that p.
Haziza has identified some important data: when we don’t know that our speakers know the answer to Q, we prefer indirect questions that ask whether they know wh-Q to direct questions that simply ask Q. And it’s awkward to ask Q immediately after asserting that one’s interlocutor does not know. Haziza’s explanation of that data is AKN. Mine is that asking direct questions when we don’t know that our speakers know is an FTA. On my view, although we have a preference for indirectly asking whether our interlocutors know (the answer to) Q rather than asking Q itself when we are unsure whether they know, this preference is defeasible and motivated by considerations of politeness rather than anything epistemic. That is at odds with AKN insofar as AKN is thought of as an epistemically inflected norm, on par with, say, the knowledge norm for assertion.
2.2. Asking the ignorant
In the last section, I argued that AKN is not the only candidate explanation of Haziza’s data. Here, I argue against AKN more directly by providing (what seem to me) counterexamples. If AKN were true, it would never be appropriate to ask questions of people when they don’t know the answer. But we routinely ask people questions even when we have no idea whether they know the answer.
I shall consider two types of cases. In the first set of cases, the primary goal of the inquirers is to get an answer to their question from the addressee. These are clearly cases of what Haziza calls “inquiry questioning.” In the second set of cases, those asking the question are primarily interested in some other goal, such as expressing their wonderment or getting their interlocutor to take up their question.
2.2.1. (Merely) hoping for an answer
Although we typically ask questions expecting an immediate answer, this is not always so. This often happens when we are surrounded by strangers and need to ask someone without much reason to choose one addressee over another. Consider the following mundane case:
Train Station
Ann: Excuse me, what time does the next train to Glasgow depart?
Bob: Sorry, I don’t know. I wish that I did.
Although Ann asks Bob something he doesn’t know the answer to, there’s no clear sense that Ann did something wrong. She’s not engaged in obviously bad conversational behavior like someone who lies or who gives information that is not well-supported by their evidence.
Haziza puts much weight on the idea that it would be better for Ann to ask the indirect question “Do you know what time the next train to Glasgow departs” (Haziza Reference Haziza2023: 16)? I agree (though I think the relevant kind of betterness has to do with politeness). But this is different from the judgment that Ann has actively done something wrong.
Indirectness is one way to redress an FTA. But it isn’t the only way. So a probative question at this stage is whether Ann could improve upon her conversation in Train Station in the same way by employing a different face-preserving strategy. I think the answer here is yes. Imagine these various ways the conversation could have gone:
Train Station*
Ann 1: Excuse me, do you know what time the next train to Glasgow departs?
Ann 2: Excuse me, I’m wondering what time the next train for Glasgow departs.
Ann 3: Excuse me, you might not know this, but what time does the next train to Glasgow depart?
Ann 4: Excuse me, there’s no reason you should know this, but what time does the next train for Glasgow depart?
Bob: Sorry, I don’t know. I wish that I did.Footnote 6
In each of these cases, Ann executes a strategy for helping Bob save face. Ann 1 employs the strategy we’ve already spent the most time with: indirectly asking Bob whether he knows. Ann 2 keeps her question “off the record,” allowing the addressee to avoid picking up the implied question if preferred (e.g., by responding “I also wonder what time it leaves − perhaps the attendant knows?”).Footnote 7 Ann 3 and Ann 4 each communicate, in different ways, that Bob is not expected to know the answer to her direct question, mitigating the loss of face if Bob is unable to provide an answer.
If AKN is the right explanation of our preference for indirect questions (when we don’t know that the addressee knows), then there should be a big difference between Ann 1 and 2 (on the one hand) and Ann 3 and 4 (on the other). That’s because Ann 1 and 2 abide by AKN − Ann 1 by asking her question indirectly, and Ann 2 by not asking a question at all − whereas Ann 3 and 4 violate AKN. But in fact, all four speeches are about equally acceptable. We find sameness where AKN predicts difference. That suggests that, to the degree that Ann’s direct question in the original Train Station case is dispreferred, it’s because she does not execute any face-preserving strategies, not because she violates AKN.
Notice, too, that the seeming propriety of Ann 3 fits uncomfortably with Haziza’s explanation of Moore-paradoxical assertions. Consider:
-
(3) I don’t know that it’s raining, but it’s raining.
is Moore-paradoxical, but so is:
-
(4) I might not know that it’s raining, but it’s raining.
Even if (4) isn’t quite as paradoxical as (3), both sound problematic. But the utterance in Anna 3 is free of the whiff of paradox:
-
(5) Excuse me, you might not know this, but what time does the next train to Glasgow depart?
This makes trouble for the contention that Moore-paradoxical askings are paradoxical in the same way that Moore-paradoxical assertions are.
Let’s consider one final twist on Train Station. In this version of the story, Ann doesn’t ask a random fellow passenger; Ann asks the customer service agent. But in this version of the story, Ann knows (or at any rate, reasonably assumes) that this isn’t something the customer service operative knows off the top of their head. Rather, Ann expects that they will look up what time the next train leaves for Glasgow on their computer.
In this case (if you like, call it Train Station**), the customer service agent is in a position to know what the answer is in the sense that they can learn the answer to Ann’s question by looking it up. But there’s no pressure in this case to say that Ann should have instead asked the indirect question “Do you know when the next train to Glasgow departs?” After all, Ann already knows the answer to that question is no − it’s assumed the agent will have to look the information up. Moreover, it’s the job of the customer service agent to answer queries like Ann’s. And this social arrangement makes it hard to generate the feeling that Ann is being impolite by asking the direct question, even if the agent doesn’t know the information offhand and needs to look it up.
In Train Station**, Ann still expects a more or less immediate answer to her question. So one could push back by arguing that there’s a sense in which, for all practical purposes, the customer service agent counts as close enough to knowing when the next train leaves.Footnote 8
But once we open the door to asking people questions they are merely in a position to know, it’s not hard to stretch out the cases so that the addressees are further and further from knowing. Consider this case:
General and Scout: A general sends a telegram to a team of scouts behind enemy lines. The general wants to know how many soldiers the enemy has assembled in their camp to the southeast. The general knows it will take his scouts a couple of days to get to the enemy camp, count the troops, and report back. The telegram reads simply: “How many troops are in the enemy’s camp?”
The scouts are (or at least may be) in a position to (eventually) know how many troops are in the enemy’s camp. But it seems wrong to say that the scouts are close enough to knowing how many troops are in the enemy camp for practical purposes. The general is sending them on a potentially dangerous mission to find out that information − it’s not information they have at their fingertips.
Once again, notice there is no pressure for the general to ask his scouts the question indirectly. The general’s social position is one that allows him to be direct with his troops without impropriety. Indeed, if the general had asked the indirect question “Do you know how many troops are in the enemy camp?”, it’s not clear that his scouts would have interpreted this as implicating the corresponding direct question. Plausibly, that’s because the general’s social position of authority removes (or at any rate weakens) the pressure to be indirect: if the general had wanted to know how many troops were in the camp, he could have just asked that. This, again, suggests that our preference for indirect questions when our addressees may not know is explained by rules of manner, not rules of epistemology.
So far, we’ve mainly focused on cases where the addressee is presumed to be in some kind of position to know the candidate answer to our question. But not all cases are like this. Sometimes, we exploit the fact that we can get information from non-answers to questions. In such cases, we can ask without expecting our addressee to (even be in a position to) know. Consider the following case:
Sleeping: Ann and Bob are watching their favorite tv show before bed. Bob’s eyes are half-closed, his mouth is slightly open, and Ann hears a noise that might have been a snore. Ann asks Bob: “Are you asleep?”
Sorensen (Reference Sorensen1981: 155–57) calls this a “pragmatic paradox liable interrogative.” What he means by this, roughly, is that one of the possible direct answers to the question is such that it cannot be properly asserted by the addressee. In this case, if Bob is asleep, he cannot properly assert “I am asleep” (even though he can properly assert “I’m not asleep!” if he’s awake).Footnote 9
Ann’s question is a perfectly sensible one to ask. And yet, precisely because she is unsure whether Bob is asleep, she is also unsure whether Bob knows that he is asleep − at least on the assumption that people who are asleep do not always know that they are asleep.Footnote 10 So this, again, is an example of properly asking a question without knowing that one’s addressee knows the answer.
Finally, self-addressed questions pose a special challenge for AKN. One can assert to oneself (as in self-encouragement), one can issue commands oneself (“Get it together, Chris”), and one can ask questions of oneself. Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 159) give this example of a self-addressing question:
-
(6) Now where did I put my wallet?
After asking myself this question, I may get up and go look, thereby (in the good case) coming to know the answer to my question. Self-addressed questions are particularly tricky for proponents of AKN like Haziza (Reference Haziza2023: 1) who also endorse the ignorance norm, according to which a speaker should not know the answer to the question they ask. Obviously, in the case of a self-addressed question, the speaker cannot be ignorant of and know the answer to their question simultaneously. So given the propriety of some self-addressing questions, AKN and the ignorance norm cannot both be true.Footnote 11
2.2.2. Asking without (much) hope for an (immediate) answer
All the examples of the last section are paradigmatic instances of what Haziza calls “inquiry questioning,” where the speaker is genuinely inquiring into the question they ask and trying to get the answer by asking. Such inquiry questionings are paradigmatic of the speech act of asking questions. But not all askings are paradigmatic.
Haziza’s version of AKN applies only to inquiry questionings, and so it’s reasonable to ask whether the future cases in this section really challenge AKN as Haziza understands it. It will be easier to evaluate this issue after examining the cases. But I briefly flag that nothing in these cases requires that the agents aren’t genuinely curious about the answer to the questions they ask or that they wouldn’t be happy to receive a knowledgeable answer from their interlocutor were it to be offered. It’s just that getting an immediate answer isn’t the primary purpose of the speech act.
Usually, we ask a question hoping for an answer from the addressee. But sometimes, asking a question sets a topic, but without the implication that the addressee will (necessarily) be the one to answer it. Consider the following case:
Team of Detectives : DCI Morse begins working with a team of detectives on a murder case after Prof. McDiarmid was killed at a gala at Lonsdale College. In order to figure out whodunnit, and immediately upon arriving at the scene of the crime, the team tries to establish a list of potential suspects based on who was at the gala. It’s hard to determine who exactly was at the gala at the time of the murder, and Morse decides that this strategy is getting them nowhere. “We need a different approach,” Morse says. “We should try to figure out if anyone benefits from her death. Did she have a life insurance policy, perhaps?”
This case plausibly exemplifies what Habgood-Coote (Reference Habgood-Cooteforthcoming: §4) has in mind when articulating the category of a shared asking, within his broader argument that askings are proposals in shared inquiry. Morse is setting the questions that he wants his team of detectives (including himself) to focus on. He certainly isn’t expecting an answer right then. And it’s not clear he expects an answer from the other members of his team − maybe Morse will be the one who discovers whether Prof. McDiarmid had a life insurance policy. Even so, his questions seem sensible. His asking them succeeds at setting strategic lines of inquiry, even if some of those lines of inquiry turn out to be dead ends.
Morse’s question is plausibly also an example of what Millson and Risjord call topic-giving questions (Millson and Risjord Reference Millson, Risjord, Stovall and Koren2025: 159). We often ask topic-giving questions when teaching. And since philosophical questions can be difficult, we often ask them without expecting a (knowledgeable) answer from our students.
Philosophy Professor : Prof. Myrtle Willoughby is teaching an introductory epistemology course. She’s been leading her class through the classic Ford and Nogot Gettier cases, and her class is handling them well. She decides to challenge them with a harder and more controversial case: “Suppose you wind up in fake barn country…” she begins. Do you know there’s a barn there when you happen to be looking at a real one?” Prof. Willoughby doesn’t expect the class to settle the question—she’s unsure of the answer and still wonders about the question herself! But she does want to get her class thinking about the fake barn case for pedagogical purposes. She wants them to join her in wondering whether the protagonists in such cases have knowledge.
Plausibly, this is the sort of case Bell (Reference Bell1975: 209) has in mind when he writes that a question can be raised without being put: “a lecturer may raise a question … simply leaving his audience to consider it if they wish.” Notice that Prof. Willoughby’s question isn’t what is sometimes called an exam question.Footnote 12 Her goal in asking the question is not to test how much her student knows. Nor is she in any sense pretending to be ignorant (she really doesn’t know what to say about fake barn country!). Her question reflects something she is really wondering about. If one of her students were able to provide a knowledgeable answer, their response wouldn’t be out of place. Even so, she asks it without expecting a knowledgeable answer from her students.Footnote 13
In concert with Bell’s distinction, Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 160) argue that topic-giving askings like those of DCI Morse and Prof. Willoughby “announce a line of inquiry,” whereby the speaker “acknowledges an erotetic commitment without making others apokritically responsible.” By “not making others apokritically responsible,” they mean (roughly) not imposing a defeasible obligation on the hearer to provide the answer to the question (Millson and Risjord Reference Millson, Risjord, Stovall and Koren2025: 152–54). If asking a question can be a way of merely raising the question without expecting an answer to it, then we can do so without abiding by AKN.
Topic-giving questions aim to get one’s audience to focus on a question rather than to give an answer. Certain other askings are not obviously aimed at getting our addressees to do anything; rather, they express our desire to know. We might call these expressive questions. This passage from van Elswyk (Reference van Elswyk2025: 262) is instructive:
Consider uses of interrogatives in contexts where the speaker knows or believes that the other conversational participants do not have the answer(s) to the question that is asked. Or, relatedly, consider questions that a speaker poses to themselves, such as those asked during a soliloquy. These are admittedly atypical uses. However, there is nothing defective about them as askings, the default speech act performed by using an interrogative. In these cases, it seems that the speaker merely expresses the desire to know the answer(s). They do not expect this desire to be satisfied because the addressee will tell them.
Such questions often manifest a sense of wonder or lament. Asking a magician “How did you do that‽”, knowing that the magician will not reveal their secrets, still succeeds at expressing what one wants to know. “How did she do that‽”, addressed to the equally befuddled audience member at your side, asks a question whose answer the addressee does not know.
Or Juliet famously asks, “Wherefore art thou Romeo?” When Juliet asks this, she expresses a desire to know something − why has she, a Capulet, been fated to love a Montague? But although she may be genuinely curious, she does not expect an answer. Nor is Romeo any better positioned to answer it.Footnote 14 Her question does not anticipate an answer. But it is still a good asking, in part because it succeeds at expressing the questions she is gripped by.
Finally, we can sometimes ask questions even when it is logically impossible that anyone should ever give us a knowledgeable answer. For some truths are (known to be) unknowable, and yet it is sensible to ask them. Suppose no one will ever know whether the total history of the universe contains an even or odd number of stars (space and the total history of the universe are just too large). Even so, here’s a question I could sensibly ask:
Stars: Is it the case that (i) the total history of the universe contains an odd number of stars, but no one will ever know that, or is it instead the case that (2) the total history of the universe contains an even number of stars, but no one will ever know that?
The answer to this question is in principle unknowable. That’s because each alternative (each answer) has a Moore-paradoxical form. If anyone knew (say) that <the universe contains an odd number of stars but no one will ever know that>, then they would also know that the universe contains an odd number of stars. And they would know that no one will ever know that the universe contains an odd number of stars. But then it would be false (and so they wouldn’t really know) that no one ever knows that the universe contains an odd number of stars. So supposing that someone knows an answer to the question gets us into a contradiction.Footnote 15
But even though this question cannot be asked in a way that satisfies AKN, Stars does not feel paradoxical in the way that a Moorean assertion does. It is of course unreasonable to expect an answer Stars, but it isn’t paradoxical merely to raise the question by asking it.Footnote 16
Stars is special because it is constructed to be metaphysically impossible to knowledgeably answer. But, in fact, I think I suspect we ask unanswerable questions with dreary frequency. Counterfactual questions are sometimes like this. We wonder what would’ve happened if we had gone to a different college or had not ended that relationship or had accepted that job offer or… I think there’s good reason to be skeptical that anyone will ever really know the answer to those questions, but it’s certainly something we can wonder about and ask aloud. Some philosophers have famously thought that no one will ever know whether there is a God. Maybe they are right, and maybe they are wrong. But we wouldn’t expect people to stop wondering or asking whether there is a God even if the question turns out to be impossible to (knowledgeably) answer.
2.2.3. The scope of a norm
Admittedly, some of the askings in the last subsection have been less than paradigmatic insofar as they do not anticipate an immediate, knowledgeable response from the addressee. Haziza might object that these cases are not responsive to his version of AKN, since his version is explicitly restricted to “inquiry questioning,” in which the speaker is inquiring into and trying to settle the question they ask by so asking.
Haziza’s strategy suggests he thinks of AKN as applying to a subspecies of asking. Others have suggested that asking a question without trying to get an answer from one’s interlocutor isn’t really asking at all. For example, Flamini (Reference Flamini2026a: 25) claims that “one asks a real question by uttering an interrogative sentence only if one’s utterance is an attempt that aims to get the addressee to answer the question it expresses.”Footnote 17 Merely topic-giving and expressive questions fall short of “real questions” on Flamini’s proposal.
Both of these strategies would ease the pressure from the purported counterexamples in the last subsection. But I think there’s a theoretical cost to each of them. It’s both plausible and theoretically simpler to suppose that all the examples of the last two subsections belong to a common speech act type and are subject to a unified norm.
There may be an important sense in which the default aim of the activity of asking a question is to elicit the answer from one’s addressee, but we shouldn’t infer that each real asking of a question thereby involves an attempt by the speaker to elicit the answer. Even if the primary function of asking questions as a practice is to elicit a knowledgeable answer from one’s interlocutor, not every instance of asking has to satisfy this function in order to be norm-abiding. This is a familiar lesson from the literature on assertion. Even if the reason we have a speech act of assertion is to allow us to exchange information, speakers can permissibly use assertion to accomplish other aims, e.g., to encourage people by reminding them of what they already know, or to get one’s opinion on the record even when (you know that) it won’t persuade anyone listening.Footnote 18 The raison d’être and licensing conditions for a speech act can come apart.
Finally, even if topic-giving and expressive questions often have a goal other than immediately settling the asked question, that doesn’t mean they are not also asked with the (perhaps ultimate) goal of resolving an inquiry into the question. Morse doesn’t expect an immediate answer to his question, but he certainly hopes to resolve the question and eventually solve the murder. And he expects his topic-giving questions to be instrumental in ultimately answering the questions that he has. Prof. Willoughby might not have much hope that one of her students will presently solve the Gettier problem. But we needn’t suppose that she doesn’t really want to know the answer to her question or doesn’t have the goal of resolving it. That is to say: as long as a speech act can have two goals (like any act), we can imagine both topic-giving and expressive questions as also meeting Haziza’s criteria for inquiry questioning.
2.3. AKN reconsidered
We’ve considered a wide range of cases that jointly suggest it is often perfectly fine to ask someone a question even if we don’t know that they know the answer. Admittedly, we often ask questions because we want (more or less) immediate answers.Footnote 19 When that’s the case, it’s a good idea to ask those who (we know) already know. But this isn’t always the case.
Direct answers are the paradigm helpful response to a question. But direct answers aren’t the only helpful responses. Partial answers (p or q), hedges (probably p), and hypotheses (perhaps p) can all be helpful responses to a question. Sometimes we helpfully respond to questions with other questions, by asking follow-up questions that would (if answered) help resolve the first.Footnote 20 AKN needlessly prohibits us from seeking out these motley responses when a direct answer is unavailable.
Sometimes, the best option available to us for getting the answer to our question is to ask someone who doesn’t (yet) know. (Train Station** and General and Scout are like this). And sometimes, our purpose in asking the question is not (merely) to get an answer, but to spread a question. Topic-giving questions like those from our detective and philosopher fit this model. Sometimes, like Juliette, we ask a question in order to express what we’re curious about. In most of these cases, it’s not just that the addressees didn’t happen to know the answer to the question they were asked. The speaker knew full well (or at least thought it likely enough) that the addressee did not know. So these cases are not merely excused or otherwise blameless violations of AKN − they are real challenges to it.
Topic-giving questions, in particular, raise an important issue. If we sometimes ask questions in order to transmit them and not (merely) to get an answer, when is it proper to do so? Are we entitled to spread questions whenever we like on whatever topic interests us? Or, as is the case with testimony, are there standards and norms for when it is appropriate to put a question forward? Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 161) suggest the latter:
Topic-giving inquiries sometimes invite others to join in the task of answering a question. In these cases, the speaker is not only acknowledging an erotetic commitment; they’re also licensing others to do so. They are, at least implicitly, attributing to the audience an entitlement to adopt the same interrogative attitude they’ve expressed.
These are the issues to which we now turn.
3. Norms for asking
3.1. Theses and terms
Although I’ve been critical of AKN to this point, I actually think it’s not far from the truth. We can think of AKN as having two related requirements. First, those who ask a question must know their question has an answer. Second, those who ask a question must know that their addressee knows the answer.
I think that the second of these requirements is too strong. But I endorse the first. One should only ask a question if one knows there is a direct answer to the question:
Knowledge Norm: One ought to: Ask Q (unconditionally) only if one knows that Q has a true, direct answer.
This is a strong norm. Notice that it also entails the (also controversial) soundness norm:
Soundness Norm: One ought to: Ask Q (unconditionally) only if Q has a true, direct answer.
Variants of this norm have been defended as a norm of IAs in Willard-Kyle (Reference Willard-Kyle2023), Whitcomb and Millson (Reference Whitcomb and Millson2024), Carter and Willard-Kyle (Reference Carter and Willard-Kyle2025), and Willard-Kyle et al. (Reference Willard-Kyle, Millson and Whitcomb2026) and of inquiry in Rosa (Reference Rosa2024). In this section, I argue that this norm applies equally to askings. In this, it goes further along a path begun by Borge (Reference Borge, Sbisà and Turner2013) who argued that we should truly believe the presuppositions of our asked questions, and it partially agrees with a recent paper by Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 149, 156), who argue that those asked questions that acknowledge an erotetic commitment of the speaker require agents to be in a position to know that they are sound.Footnote 21
I also think it’s plausible − although I will not argue for it here − that there is an ignorance norm for asking:
Ignorance Norm: One ought to: Ask Q (unconditionally) only if one doesn’t know a direct answer to Q (as an answer to Q).Footnote 22
This norm has been defended explicitly for (a certain kind of) asking by Whitcomb (Reference Whitcomb2017) and as the default norm for questions expressed with the interrogative mood by Gaszczyk (Reference Gaszczyk2026). There is broader support for an ignorance norm that applies either to inquiry or to IAs, a thesis defended also in Whitcomb (Reference Whitcomb2010), Friedman (Reference Friedman2017), Millson (Reference Millson2021), Sapir and van Elswyk (Reference Sapir and van Elswyk2021), Rosa (Reference Rosa2024), and Willard-Kyle et al. (Reference Willard-Kyle, Millson and Whitcomb2026).Footnote 23 My focus, however, will be on the knowledge norm.
Before going further, it will be useful to clarify some terminology. Since Hamblin (Reference Hamblin1973), philosophers have thought that whatever questions are, they are intimately associated with the set of their candidate answers. Perhaps the question just is the set of answers. Perhaps it is a partition that imposes a set of answers. Perhaps it is something else that has a set of answers as its alternatives. No matter what, there is some close relationship between a question and its (fully resolving) answers. For example, the question <Is it raining?> can be thought of as (intimately associated with) the set of answers {it is raining, it is not raining}. The question <Will the Padres or Dodgers win the game?> can be thought of as (intimately associated with) the set of answers {the Padres will win, the Dodgers will win}. The question <Who will be the next president of the United States?> can be thought of as (intimately associated with) the potentially infinite set of answers {Marco Rubio will be the next U.S. president, Alexandria Ocasio-Cortez will be the next U.S. president, …}.
Those answers that fully resolve a question are sometimes called its direct answers (some prefer the term “complete”).Footnote 24 Not every answer to a question fully resolves it. For example, <Chris won’t be the next U.S. president> does answer (to an extent) the question of who the next president will be. But it doesn’t provide all the information that the question calls for. Similarly, the answers <The next U.S. President will be a Democrat> or <The next U.S. President will be either Marco Rubio or J.D. Vance> do not fully − but only partially − answer the relevant question. They are not direct answers. One test for whether an answer is direct is whether an agent knowing that proposition would (other things being equal) make it true to say that the agent knows wh-Q.Footnote 25 For example, if I know that it is raining, then I (typically) count as knowing whether it is raining. So that it is raining is a direct answer to whether it is raining.
Having a true, direct answer is an important property for a question to have. This property is sometimes called soundness: a question is sound iff it has a true, direct answer. Semanticists talk about the soundness of a question as its basic presupposition: The (semantic) presuppositions of a question are whatever is true given any of its direct answers.
Consider again the cases presented as objections to AKN. While none of the agents knew that their interlocutor knew the answer to their question, they all knew that their question had a true, direct answer. Ann knows, presumably, that there is a next train to Glasgow, she’s just not sure when it departs. The General knows there is some number of troops in the enemy camp. Morse knows that Prof. McDiarmid either did or did not have a life insurance policy. The epistemology professor knows that there’s an answer as to whether you know in fake barn cases, even if they aren’t sure they’ll ever know what that answer is. There is an answer to the Stars question even though no one will ever know it. This suggests that we try to ask questions that we know to be sound even if we don’t always address them to interlocutors who know the answer.
With a working conception of soundness and direct answers in hand, we’re now ready to appreciate three arguments for the knowledge norm. The first is a recycled argument originally given for the knowledge norm for inquiry. The second is a new argument from rhetorical questions. And the third is a new argument based on deception that takes inspiration from a deception-based argument for the ignorance norm put forward by Whitcomb (Reference Whitcomb2017).
3.2. Three arguments
3.2.1. The argument from possibly defective questions
The first argument comes from the seeming impropriety of asking questions that, for all we know, are unsound − questions that are possibly defective.Footnote 26 Consider Jack, Jill, and John. Suppose I know that (exactly) one person in this group loves exactly one of the others, and that’s the only relevant thing I know about them. It would be strange, in this case, to ask whether Jack loves Jill or (rather) whether Jill loves Jack. That leaves out options that (given what I know) are just as good as the others. For example, it leaves out the possibilities that John loves Jill or that Jack loves John. So, I shouldn’t ask:
-
(7) Does Jack love Jill or (rather) does Jill love Jack?Footnote 27
But I may ask:
-
(8) Does Jack love Jill, or does Jack love John, or does Jill love Jack, or does Jill love John, or does John love Jack, or does John love Jill?
Or since that’s a mouthful, I’d probably ask:
-
(9) Who among Jack, Jill, and John loves which one of the others?
The knowledge norm offers an explanation for this pattern. It’s wrong to ask (7) because I don’t know that question has a true, direct answer − it wouldn’t, for example, if (only) John loved (only) Jack. But I do know that (8) and (9) are sound: that’s why I’m epistemically entitled to ask them.
Or suppose it’s the semifinals at the World Cup, and the four teams that remain are France, Spain, England, and Argentina. At that point in time, it would be strange for me to ask:
-
(10) Will France or (rather) Spain win the World Cup?
That question may turn out to be sound (suppose Spain will in fact win). Even so, it’s a strange question for me to ask given that, for all I know, the winner could also be Argentina or England. When I ask this question, I don’t know that it is sound. That explains why it is a bad question to ask.Footnote 28
It seems wrong to ask possibly defective questions − questions that we do not know to be sound. It’s evidence for the knowledge norm that it explains this judgment.
3.2.2. The argument from rhetorical questions
The argument of the last section echoes arguments that have been developed for a knowledge norm on inquiry or IAs. In this section, I want to consider a new argument for a knowledge norm for asking, or at least for this weaker principle:
Soundness Norm: One ought to: Ask Q (unconditionally) only if Q has a true, direct answer.
The soundness norm is controversial enough in its own right. But its truth would also provide evidence for the knowledge norm insofar as the knowledge norm entails it.
This new argument is built on rhetorical questions and on the observation that we seem to be able to produce a rhetorical effect from asking flagrantly unsound questions. Rhetorical questions are often set aside when discussing inquiry-related norms. After all, aren’t rhetorical questions just assertions in disguise?Footnote 29 But even if rhetorical questions are (usually) best understood as assertions in disguise, they get their assertoric effect by posing a question. That is to say, they are indirect speech acts (Searle Reference Searle, Cole and Morgan1975, Asher and Lascarides Reference Asher and Lascarides2001) that simultaneously play the role of questioning and asserting.Footnote 30 Accordingly, (I’ll suggest) the norms governing asking questions play a role in explaining why rhetorical questions have the effect they do.
We’ll work our way back to the soundness norm, but let’s first remind ourselves of the ignorance norm for asking (Whitcomb Reference Whitcomb2017):
Ignorance Norm: One ought to: Ask Q (unconditionally) only if one doesn’t know a direct answer to Q.
At first glance, this norm runs into difficulty explaining rhetorical questions.
-
(11) A: [overhearing groans from the tv room during baseball season] Are the White Sox losing?
-
• B: Is the Pope Catholic?
In response to A’s question, B asks whether the Pope is Catholic. But surely B knows as well as anyone that the Pope is Catholic, so it looks like B is violating the ignorance norm. But there’s nothing wrong with B’s response. That’s not what the ignorance norm seems to predict.
Whereas rhetorical questions look like a problem for the ignorance norm at first glance, they are actually, on reflection, further evidence for it. And that’s because one kind of evidence for the existence of a norm is evidence that it can be flouted.
In what sense do rhetorical questions flout the ignorance norm? Sadock and Zwicky (Reference Sadock, Zwicky and Shopen1985) define rhetorical questions as “those a speaker uses just for rhetorical effect when not only is he sure of the answer, but he judges that the listener knows that he is sure of the answer as well” (Sadock and Zwicky Reference Sadock, Zwicky and Shopen1985: 180). Rohde (Reference Rohde2006: 147) says that “rhetorical questions require …obvious answer(s) in the existing commitments of the discourse participants.” Often theorists require the relevant “effect” to be the same as an assertion. For example, König and Siemund (Reference König, Siemund and Shopen2007: 283) gloss a rhetorical question as “a statement expressed by an interrogative sentence.”Footnote 31 In short, rhetorical questions typically flout the ignorance norm by asking a question whose answer is common knowledge in order to generate an assertoric effect.
When someone asks whether the Pope is Catholic, it’s (near enough) common knowledge between the speaker and hearer that the speaker knows the answer to their question. The speaker is breaking the ignorance norm and not trying to hide it. That raises the question what the speaker is up to and creates “interpretive resistance” (Whitcomb Reference Whitcomb2017: 150). They aren’t seeking to be informed about a question they don’t know the answer to, so what are they doing? We search for alternative explanations. And especially since this question is given in response to a question, it’s natural to interpret it as an answer: yes, the White Sox are losing (with the tongue-in-cheek suggestion that this fact is just as obvious as the fact that the Pope is Catholic).
It would be wrong, therefore, to conclude that rhetorical questions are evidence against the ignorance norm, or even that they are irrelevant. In fact, rhetorical questions are evidence for the ignorance norm. For the fact that we flout the norm plausibly plays a role in the explanation of why rhetorical questions function the way they do. And you can’t flout a norm that doesn’t exist.
The literature on rhetorical questions has tended to focus on examples like “Is the Pope Catholic?” and “Who likes being criticized?” (König and Siemund Reference König, Siemund and Shopen2007: 283) in which the speaker and audience obviously know the question’s direct answer. But interestingly, not all rhetorical questions are like that. There are at least two ways in which some rhetorical questions depart from the paradigm.
First, the thing that’s common knowledge is sometimes less than a direct answer. Consider King Théoden’s (much memeified) rhetorical question from Peter Jackson’s rendition of The Lord of the Rings:
-
(12) Where was Gondor when the Westfold fell?
In context, Aragorn has suggested to Théoden of Rohan that he call to Gondor, Rohan’s ancient ally, requesting military aid. The implied answer to Théoden’s question is “not here” or “not standing alongside Rohan.” This is an eliminative (or perhaps partial) answer to Théoden’s question, but it isn’t a direct one: a direct answer would positively specify where Gondor’s armies were. But in this case, the eliminative answer is resolving enough given the practical aims of the conversation, which is to decide whether to send riders requesting military aid to Gondor. No one in the movie theater needs to know the direct answer to appreciate the rhetorical effect of the question.
But some rhetorical questions seem not to have a true answer at all. Indeed, some rhetorical questions seem to flout the norm that a question be sound. Here’s an example from The Sound of Music. At the beginning of the movie, Maria doesn’t fit in at the abbey. While debating what to do with her, the nuns ask (in song, of course), “How do you solve a problem like Maria?” The song includes the following exchange between the sometimes exasperated sisters and the wise Mother Abbess:
Sister Margaretta: But how do you make her [Maria] stay? And listen to all you say?
Mother Abbess: How do you keep a wave upon the sand?
Sister Margaretta: How do you solve a problem like Maria?
Mother Abbess: How do you hold a moonbeam in your hand?
The Mother Abbess’s questions are rhetorical. No one expects an answer to them. But it’s not because there is some direct answer she is ignorant of.Footnote 32 After all, no one − not even the Mother Abbess − knows how to hold a moonbeam in their hand.
But the Mother Abbess is flouting a norm governing asking: she’s flouting the norm that asked questions ought to be sound. Her rhetorical question implicates that, just as there’s no way to hold a moonbeam in your hand, so there’s no way to solve a problem like Maria. Perhaps that’s because Maria is unsolvable; perhaps it’s because she isn’t a problem to be solved in the first place. In this way, the Mother Abbess’s question has assertoric effect.
Nor is the Mother Abbess alone. Consider the question
-
(13) Why (should I bother to) be moral?Footnote 33
when asked not by the sincere meta-ethicist but by the flippant scoundrel. The scoundrel’s question gets its rhetorical effect because we judge that the scoundrel believes there is no true answer to that question. Or again, consider:
-
(14) Why not us?
-
(15) Why couldn’t it be us?
These questions, when asked defiantly by the written-off underdogs, rhetorically suggest that there’s no reason it couldn’t be them − something which can only be true just in case there is no true answer to them.
I think that once we see rhetorical questions as (often) involving flouting norms on querying generally (rather than just the ignorance norm), other examples of rhetorical questions start coming to mind. We can, plausibly, get rhetorical questions by violating other (not necessarily epistemic) norms on questioning. Imagine the following responses to 16(a):
-
(16) (a) Are you pregnant?
(b) Oh, and how much do you weigh?
(c) And how much money do you make?
(d) Have you had a vasectomy?
By asking 16(b), 16(c), or 16(d), the speaker flouts a norm on asking invasive or impolite questions. But they do so in a way that points out that the query in 16(a) violated the very same social norm, thereby explaining their refusal to answer.Footnote 34
Some readers, I suspect, will be put off by my liberal usage of “rhetorical question” as applied to all of the questions above, preferring to restrict the term to questions whose answers are common knowledge, as Sadock and Zwicky (Reference Sadock, Zwicky and Shopen1985: 180) require. I think my liberal use is defensible. The speaker doesn’t expect an answer to any of the questions above, and they all at least implicate propositional content.
But I don’t think the argument hangs on terminology. Let us say that the examples of this section have a rhetorical effect (even if they aren’t rhetorical questions, narrowly construed). They are (at least) structurally similar to rhetorical questions in that they have a rhetorical effect other than prompting an answer, they implicate propositional content, and they do that in part by flouting the norms that generally govern acts of querying. We’ve discovered that flagrant violations of a soundness norm (like the Mother Abbess’s question) can generate this kind of rhetorical effect.
Do all ways of flagrantly violating norms on asking questions generate rhetorical questions? No. As Searle (Reference Searle, Cole and Morgan1975: 70) points out, “an indirect speech act utterance of Can you reach the salt? may be defective in the sense that S may already know the answer” (obviously yes, that’s why I asked you!). In cases like this, “the speaker issues a directive BY WAY OF asking a question” (Searle Reference Searle, Cole and Morgan1975: 70) − so the speaker really is asking a question, even if the primary illocutionary effect of their utterance is directive. Even so, the question “Can you reach the salt?” is not rhetorical. Unlike rhetorical questions, the utterance anticipates a response (e.g., “yes, of course”). And unlike rhetorical questions, the utterance constitutes a directive rather than an assertion. Notably for this discussion, however, although either speech act may intentionally break the ignorance norm, the breaking of a norm plays an essential role in the rhetorical effect of Is the Pope Catholic in a way that it doesn’t in Can you reach the salt. Asking whether the Pope is Catholic in a room of people who have never heard of the Pope would fall flat as a rhetorical question. But I may succeed at getting you to pass me the salt by asking whether you can reach it even if the salt is just far enough from you that I’m really unsure as to the answer.
Let’s take a step back. What argument does this section suggest?
We get rhetorical (or rhetorical-like) questions, in part, by flouting norms on asking questions. (e.g., “Is the Pope Catholic?” flouts the ignorance norm.) But not all rhetorical questions flout the ignorance norm. Some questions like the Mother Abbess’ “How do you hold a moonbeam in your hand?” are better understood as flouting a soundness norm. One can only flout a norm that actually exists. Therefore, we should think there is a soundness norm for asking questions.
3.2.3. Interlude: An olive branch for AKN
I spent the first part of the paper arguing against the AKN. Now, I present an olive branch. The argument from rhetorical questions also provides evidence for AKN. That’s because we sometimes get a rhetorical effect by asking questions our addressee obviously doesn’t know the answer to.
Consider Job. Job demands an account from God as to why he has suffered despite his righteousness − declaring that he will defend himself in divine court against his apparent indictment.Footnote 35 When God responds to Job, he does so with a series of rhetorical questions. Most of the rhetorical questions are typical, in that the direct answer is common knowledge between God and Job:
-
(17) Have you entered the springs of the sea or walked in the recesses of the deep? Have the gates of death been revealed to you? (implied answer: no)Footnote 36
Some plausibly (are known to) have no answer:
-
(18) Where were you when I laid the foundation of the earth? (There is no place Job was when God laid the foundations of the earth.)Footnote 37
But others get their effect by calling attention to the fact that Job, the addressee, is ignorant of the answer:
-
(19) Where is the way to the dwelling of light, and where is the place of darkness,
-
(20) that you may take it to its territory and that you may discern the paths to its home? Surely you know, for you were born then, and the number of your days is great!Footnote 38
It’s common knowledge between God and Job that Job is ignorant of the answer to these questions. These questions get their rhetorical effect by violating the typical social expectation that the addressee (Job) knows the answer to the question. This, then, is evidence for AKN.
We’ve already seen, however, that questions can generate a rhetorical effect by violating norms of politeness (e.g., “Have you had a vasectomy?” in response to “Are you pregnant?”), which generally seek to avoid or mollify FTAs. Moreover, it is crucial to God’s questioning of Job that he is humbling him by displaying Job’s ignorance, one of the ways Brown and Levinson (Reference Brown and Levinson1987: 68) specify as a way to lose face. God’s questions to Job are FTAs. It is humbling for Job to admit the limits of his knowledge after declaring he would approach God “like a Prince”Footnote 39 in court. And indeed, Job’s final words in the book are self-effacing: “I despise myself and repent in dust and ashes.”Footnote 40 The effect of the question is not to get Job to respond with an answer but to humble Job − to cause Job to lose face.Footnote 41
What, then, shall we say about the rhetorical effect of God’s final question? Does it get its rhetorical effect from a distinctively epistemic AKN or from more general considerations favoring politeness and face-preserving speech acts? Considered independently, I think both answers are plausible. Given the apparent counterexamples to AKN in Section 2, I prefer the second alternative. Even so, we should give AKN its due: AKN offers a simple explanation of why some rhetorical questions are generated by the common knowledge that the addressee does not know the answer.
3.2.4. The argument from deceptive questions
One of Whitcomb’s (Reference Whitcomb2017: 150) arguments for the ignorance norm is that we can deceive people into thinking that we don’t know the answer to a question by asking it. Here’s one of his examples:
[S]uppose that my Dean is forming a committee to study teaching evaluations, and says that any committee members should have some knowledge of psychometrics. If I have been studying psychometrics for years but don’t want to do any committee work, I can just ask the Dean ‘What’s psychometrics?’. By asking this, I can mislead her into thinking I don’t know what psychometrics is. (Whitcomb Reference Whitcomb2017: 150)
The ignorance norm is a good explanation of the fact that we can thus deceive each other with questions. The Dean believes that the protagonist doesn’t know what psychometrics is because asking a question implies ignorance of the answer.
But we can also deceive each other by asking questions we don’t know to be sound. This analogously suggests the knowledge norm. A particularly grim example of deception through questioning can be found in police interrogations. Consider this passage from Wright (Reference Wright2026: 1005–1006):
Interrogators are encouraged to being the interrogation by claiming that they have sufficient evidence of the suspect’s guilt, …[and] interrogators redefine the nature of the interrogation, making it seem that the suspects’ guilt is already established and that they simply want to understand why the crime happened.
In effect, this is a shift from asking questions that aim to establish whether the suspect committed the crime to questions that aim to establish why the suspect committed the crime − it’s a shift from a question that does not presuppose the suspect’s guilt to a question that does.
For example, suppose an interrogator asks the following without evidence:
-
(21) How come we found your DNA at the scene of the crime?
Assuming they have found no DNA evidence, this would be a deceptive question. Why is it deceptive? Because the interrogator doesn’t actually know that the question is sound. If the suspect’s DNA is not at the scene of the crime, then the question has no true answer.
Notice that the question is deceptive even if the suspect’s DNA is at the scene of the crime. The interrogator would still falsely make it seem like they knew that the suspect’s DNA was at the scene of the crime, when in fact (we are imagining) they have no such evidence.
Indeed, the ability to deceive with questions is one of the things Belnap and Steel (Reference Belnap and Steel1976: 115–16) flagged in their pioneering work on the soundness (they would call it the “truth”) of questions:
To ask something with a false presupposition … is very much like making a false statement. One can do it knowingly and maliciously, and be exactly on a par with a liar … For this reason we hereby make it a proposal to call a question “true” or “false” according … as it does or does not have some true answer. (Belnap and Steel Reference Belnap and Steel1976: 115–6)
Intentionally asking an unsound question is like intentionally asserting a false proposition. Both are means of deception.
As wary students know, one species of the deceptive question is the trick question. Consider this familiar example:
-
(22) If a plane crashes on the border of California and Nevada, where are the survivors buried?
Of course, the survivors aren’t buried anywhere. This counts as a trick question because it is “designed to box recipients in, forcing them to choose among propositions none of which they endorse” (Millson and Risjord Reference Millson, Risjord, Stovall and Koren2025: 153). The appropriate response is not a direct answer at all, but to reject a presupposition of the question. The intentional unsoundness of the question explains why it is a trick question and why students feel justified in complaining about trick questions on exams.
Just as the ignorance norm explains how we can ask questions deceptively by implying that we are more ignorant than we are, the knowledge norm explains how we can ask questions deceptively by implying that we know more than we do.
4. Objections
I’ve argued that there is both a soundness and a knowledge norm on asking. One should only ask questions that have true, direct answers; indeed, one should only ask questions when one knows that they have true, direct answers. In this section, I consider objections. In response to the final objection, I’ll modestly weaken the norm that I ultimately defend.
4.1. Presupposition-checking questions
Habgood-Coote (Reference Habgood-Cooteforthcoming) argues that “we can use asking to check whether the presupposition of a question is correct.” As an example, he offers the following sentence:
-
(23) I don’t know how your university is structured, but are you a member of the philosophy department?
This question seems fine to ask. But arguably, the question presupposes that the university is structured in such a way that it has a philosophy department, while simultaneously denying that the speaker knows the university is structured this way. If so, that’s a problem for my account.
But I think we can stare down this purported counterexample. First, the speaker could know that the university has a philosophy department and yet truly say that they don’t know how the university is structured. Perhaps even though they know the university has a philosophy department, they don’t know what other departments there are, or how the departments are grouped within various colleges, and so on.
But maybe we can change the example to get around this problem:
-
(24) I don’t know whether your university has a philosophy department, but are you a member of the philosophy department?
I contend, however, that (23) sounds significantly worse than (22). And if we change the utterance to make it even clearer that the question presupposes that the university has a philosophy department, the sentence comes out even worse:
-
(25) I don’t know whether your university has a philosophy department, but is it the case that: (a) your university has a philosophy department and you are a member of it or (b) your university has a philosophy department and you are not a member of it?
This question clearly is problematic. This question obviously ignores the open (by the speaker’s lights) possibility that the university has no philosophy department whatsoever. The clearer it is that the sentence expresses a violation of the knowledge norm, the clearer it is that the sentence is problematic. That’s the pattern we should expect if the knowledge norm is true.
4.2. Merely justified belief that a question is sound
Falbo (Reference Falbo2024) presents several objections to the knowledge norm. While her discussion focuses on norms governing inquiry, some of her cases neatly translate into objections to a knowledge norm for asking. Here’s one of them:
Dog Park: Randy, a stranger, approaches Thomas at the dog park. He starts to pet the dog that Thomas walked into the park with. Randy asks: “What’s your dog’s name?” Thomas answers: “Porco.” Unbeknownst to Randy, the dog isn’t Thomas’s—he’s dogsitting Porco for his roommate—but Thomas doesn’t mention this because it would be pedantic and irrelevant to Randy’s reasons for inquiring. Randy just wants to know the dog’s name so he can use it while playing with the dog for a couple of minutes at the park. (Falbo Reference Falbo2024: 970)
Thomas (let’s suppose) doesn’t have a dog. Any answer to Randy’s question would have the form “Thomas’s dog’s name is x,” but every proposition of that form is false. So the question is unsound. But then Randy has violated the knowledge norm for asking. But, arguably, Randy does nothing wrong by so asking. Certainly, Randy hasn’t behaved in an epistemically irresponsible way. Ostensibly, that’s a problem for the knowledge norm.Footnote 42
Nor are these cases rare; on the contrary, they are common in the history of science. As Henne (Reference Henneforthcoming: §5) notes:
In the history of science, we find countless scientific inquiries into questions that make no sense today (Jardine Reference Jardine2000) or that rely on an incorrect or imperfect conceptual framework, e.g. by carving the possibility space in the wrong way or by incorporating incorrect assumptions about what objects and properties exist in the world. Such inquiries ask defective questions in Willard-Kyle’s sense. The fact that norms of inquiry would forbid such commonplace and widespread cases of inquiry is …suspicious.
As Henne notes, both the presuppositions and framing of historically important questions often turn out to have been unsound or otherwise problematic. And yet surely scientists have, in some sense, been asking the right questions as our knowledge of the physical world has evolved.
These cases mirror a certain kind of objection often made against the knowledge norm for assertion. Knowledge is factive, so whenever we say something false, we thereby violate the knowledge norm for assertion. But sometimes we assert things based on our best evidence and conceptual schemes or as a result of employing our typically reliable cognitive processes and, nevertheless, get things wrong. In such cases, we assert what we reasonably or reliably or rationally or justifiably believe even though we fall short of knowledge. Many people have the intuition that in such cases, we perform as we ought to, assertorically speaking. If they are right, the norm of assertion is not knowledge but something weaker and non-factive: the norm of assertion is justification, or some variant thereof.Footnote 43
This standard objection has a standard reply. Such assertions, though norm-violating, are excused (roughly) because the agents were behaving in the sort of way that generally leads to complying with the knowledge norm of assertion. So if Randy went home and told Sandy, “I met Thomas’s dog, Porco, at the dog park,” he would be asserting responsibly even though, because false, his assertion does not satisfy the knowledge norm.
Unsurprisingly, this reply has not satisfied everyone. One reason this reply has not satisfied everyone is that the debate turns on deep theoretical disagreements about the nature of norms and how externalist (or internist) we should expect them to be. We won’t make further progress on that aspect of the debate here.Footnote 44
But another reason this reply has not satisfied everyone is that it seems insufficient merely to say that Randy is not blameworthy. Rather, Randy does something good as a question-asker; merely saying that he failed to be blameworthy does not capture this positive reaction to Randy’s question. As Falbo (Reference Falbo2024: 971) puts it, it’s not that Randy is “merely blameless or excused for having inquired into Q [namely, “What is Thomas’s dog’s name?”]. Instead, they positively ought to have inquired into Q.” We need to be able to say something good about Randy’s question, not merely to abstain from one species of criticism.
I think we can, however. Notice that Randy succeeds in manifesting a disposition to succeed at asking sound questions reliably enough in normal circumstances. (In related work, Adam Carter and I have thus described inquirers like Randy as satisfying a competence norm, even if they fall short of full, erotetic success.)Footnote 45 After all, it’s normally a dog’s owner who takes them to dog parks. The disposition to ask people at the dog park what their dog’s name is normally results in asking sound questions. So, it’s not merely that we can avoid blaming Randy: we can praise him for manifesting an interrogative competence − a competence to ask questions only if they are sound reliably enough. Compatibly with that, we can deny him full interrogative success by saying that he falls short of the knowledge norm.Footnote 46 Just as we can say something good about the speaker whose assertion, though justified, falls short of knowledge − it manifest a justified belief and a disposition to assert only what they know − so we can say something good about the speaker whose question falls short of the knowledge norm in the analogous way.
4.3. Interpersonal questions
Another of Falbo’s cases concerns the care with which we should ask interpersonal questions:
Best Friends: Lucinda and Tyra have been best friends for decades; they are as close as can be. Tyra isn’t a big fan of Lucinda’s on-again-off-again boyfriend, Clay. Clay likes to drink—a lot. Over the past few months, Clay’s drunken blackouts have become more frequent and more aggressive. Tyra has started to worry about Lucinda’s safety. During a visit, Tyra notices that Lucinda isn’t her usual happy-go-lucky self. She seems tired and deflated. Tyra also notices several small finger-sized bruises scattered across Lucinda’s arms. Lucinda and Tyra make eye contact, and in that moment, Lucinda realizes that Trya has seen her bruises. Lucinda quickly crosses her arms, covers up the bruises, and nervously looks away. Assuming the worst, and feeling deeply concerned for her friend’s safety, Tyra asks: “How long has he been doing this to you?” (Falbo Reference Falbo2024, 972).
There’s a lot going on in Falbo’s case, which is carefully constructed with an eye toward skillfully navigating delicate and (perceived to be) shame-inducing topics in interpersonal inquiries.
First, let’s see how this case is a challenge to the knowledge norm for asking. We are to understand the case as one in which Tyra suspects or assumes that Clay has hit Lucinda but does not know that for sure. All answers to Tyra’s question, “How long has he been doing this to you?” have (something like) the form, “Clay has been hitting me for x period of time.” Accordingly, Tyra’s question only has a true, direct answer if Clay has been hitting Lucinda for some period of time. But since we assumed that Tyra does not know that Clay has been hitting Lucinda, neither does Tyra know that her question has a true, direct answer. And nevertheless, it seems like Tyra is asking a good question in an especially sensitive way. That’s a problem for the knowledge norm.
What’s the best response? As a first pass, one might say that this is a case in which Tyra is performing well as a friend but poorly as a querier. Clearly, in this instance, it’s more important to be a good friend than a good querier − to pay more attention to what Lucinda needs as a friend than to what the norms for asking legislate.
Even though I think there’s ultimately something right about this response, it is unsatisfying as developed so far. After all, shouldn’t the norms governing acts of questioning be sensitive to our goals as agents? And if we are relational people who thrive when we can ask each other good interpersonal questions, why shouldn’t facts about what it is good for Lucinda to ask as a friend be relevant to what it is good to ask as a questioner?
A more complete response to Falbo’s challenge is this: that Tyra is violating (or at any rate, skirting) the rules governing question-asking in part explains the way in which she succeeds in being a sensitive friend. But then, of course, we need to posit the existence of the rules Tyra is (perhaps judiciously) skirting.
Here’s the core thought. Lucinda passes over the issue of whether Clay has hit Lucinda. There are (presumably) lots of questions that Tyra does not ask Lucinda in this exchange. She doesn’t ask (we may imagine) what her plans for the weekend are, or what her mother’s middle name was, or whether she wants to get lunch. All these questions go unasked. But the question “Has Clay hit you?” is not just unasked: it is elided. Passed over. Skipped. The judgment that something important is passed over is consistent with some of the illuminating things Falbo says about the case:
[By not asking] the more direct and confrontational question (“Are you being physically abused?”), Tyra is perhaps aiming to minimize the amount of potential shame or embarrassment that Lucinda may feel. In this sense, we might interpret Tyra’s question as what Brown and Levinson (Reference Brown and Levinson1987) call a face-saving act: it is an attempt to reduce the stigma or negative impact that this conversation may have on Lucinda’s reputation or self-image.
Given that we are to understand the case as one in which Tyra really doesn’t know whether Lucinda has been abused, she’s taking a gamble by skipping over the question. If she’s not being abused, Lucinda may be upset and defensive (“Why would you assume that?”). But if she is, Tyra may guard against the (undeserved but real) shame or embarrassment that Lucinda may feel by directly admitting it. By treating the admission as past − by smuggling in the information that she is being abused into the presupposition of the question rather than making it the at-issue content − Tyra graciously passes over a question that may shame her friend.
Lucinda may be grateful that her friend has passed over this question. But in what sense is Tyra passing over the question of whether Lucinda is being abused? In order to satisfy the knowledge norm for asking with respect to the question “How long has this been happening?”, she’d have to know the abuse has been happening. Tyra counts as skipping the question “Has he been abusing you?” because she needs to know the answer to that question before she’s in a position to properly ask the question she does ask.
In this way, I think the knowledge norm is part of the explanation of the questions (asked and elided) in Good Friend. True, Tyra has not abided by the knowledge norm for asking. But she succeeds at being a good friend by skirting the rules. Tyra skips over a question that, in some sense, ought to have been asked first. As a good friend, Tyra probably made the right decision. But the feeling of relief that Lucinda feels upon having the question skipped can only be explained if there’s a sense in which the question ought to have been asked. The knowledge norm for asking is the rule that Tyra judiciously circumvents by skipping ahead.
4.4. Challenge questions
Challenge questions pose an especially tricky problem for the knowledge norm. Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 158), who are otherwise quite sympathetic to the idea that asked questions should be sound, give the following as an example of an addressee-challenging question:
-
(26) Why should I believe what you say? [Context: the interlocutor has just made an assertion]
Its apparent propriety is especially challenging for soundness-based norms since through it
“speakers seem to express a belief that there is no true answer to their question” (Millson and Risjord Reference Millson, Risjord, Stovall and Koren2025: 158). That’s because the question apparently presupposes that I should believe what you say but also (when asked challengingly) implicates that I don’t yet believe you. Since I don’t yet believe you, it’s also doubtful I think I should believe what you say, since it is baldly akratic to simultaneously think I should believe that p while not doing so. Therefore, it appears that we don’t always believe that our challenge questions have true answers − we ask information of the speaker expecting them to fail because (we judge) there is (or at least may be) no true answer to give.
Challenges also sometimes come in conditional form. In the opening scene of Star Wars, Darth Vader interrogates rebels after boarding their ship. One soldier refuses to cooperate, protesting that they are merely a consular ship on a diplomatic mission. Darth Vader asks a challenge question:
-
(27) If this is a consular ship, [then] where [on this ship] is the ambassador?
Vader is not really curious where the ambassador is: he knows there’s no ambassador on board. The question at the end of the conditional is unsound. That Vader’s question seems appropriate thus poses a problem for the knowledge norm.
Or rather, it would, if the knowledge norm were not restricted to cases where an agent asks a question unconditionally. Just as the knowledge norm of assertion does not require that the propositional consequent of every asserted conditional be true, so the knowledge norm of asking does not require that the interrogative consequent of every conditionally asked question be sound.
But Vader’s case is still worth considering: in a roundabout way, Vader’s question is actually yet more evidence that we expect questions to be sound. Imagine Vader had instead responded with this conditional:
-
(28) If this were a consular ship, then there would be an ambassador on board.
We would then immediately recognize Vader’s response as an enthymeme. The missing, implicit premise is that there is no ambassador on board, whereby we infer that (according to Vader, at least) this is not a consular ship.
Vader’s conditional challenge question has a similar, enthymematic feel. Whereas the missing premise in (27) is that the consequent is false, the missing premise in (26) is that the question in the consequent is unsound. In both cases, we infer that (Vader thinks that) the antecedent is false − this is not a consular ship. But if anything, this is yet more evidence that (as Belnap and Steel proposed) we treat soundness for questions like truth for propositions.
But what about the first kind of challenge question that Millson and Risjord raise − cases like “Why should I believe what you say?” − where the implication is that there’s no reason I should believe what you say? These are hard cases; even so, I think we can justify holding onto the knowledge norm and staring down the apparent counterexamples.
Challenging a speaker’s assertion is interpersonally risky: it’s an FTA for the asserter. This seems to explain why we often present our challenges as weaker than we really mean them:
-
(29) (a) The Democrats will take the Senate in the upcoming election.
(b) No, that’s not true.
(c) I don’t think that’s true.
(d) Are you sure?
(e) Why do you think that?
(f) Why should I believe what you say?
Any of these responses to (28a) is a way to challenge the assertion. But they don’t all do so as baldly as the straight denial in (28b). (29c) hedges the denial.Footnote 47 (28d) gives the speaker a chance to weaken the assertoric force of their own assertion. The rest give the speaker an opportunity to give further reasons for their assertion, suggesting that the speaker isn’t convinced yet. And the final one actually presupposes that the listener should be convinced − there is a reason to believe what the speaker says − but simultaneously asks for further explanation as to the reasons.
For this reason, I think challenge questions like “Why should I believe what you say?” or “How do you know that?” are usually best understood as face-saving pretences. It’s a strategy for questioning the speaker’s reasons without directly calling into question the content of the assertion. But because there’s an element of pretence about them − or, at any rate, an element of provisional deference − we shouldn’t build our norm to accommodate these cases.Footnote 48
4.5. Selfless questions
A final objection to the knowledge norm for asking is inspired by Lackey’s selfless assertions. Consider the following (lightly abridged) case from Lackey (Reference Lackey2007: 598–99):
Distraught Doctor: Sebastian is a pediatrician who recognizes that 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 diagnosed with autism. The surprise brought on by his daughter’s recent diagnosis cause him to abandon his previously deeply held beliefs regarding vaccines. While 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 own, recent, personal history and that he has an obligation to his patients to present what is most likely to be true, Sebastian asserts, ‘There is no connection between vaccines and autism,’ although it’s false that Sebastian himself believes or knows this proposition.
Selfless assertions are controversial,Footnote 49 but I’m sympathetic to the lesson Lackey draws from them. Sometimes, we can do well as asserters even when we can’t do well as believers. Ideally, our assertions will be sincere in the sense that we’ll believe what we say. But especially when there are psychological obstacles to believing as we ought, it’s more important that we assert what we are in a position to knowFootnote 50 than that we preserve a perfect match between our assertions and our beliefs.Footnote 51
Put another way, when we assert insincerely, something has gone wrong. But the thing that has gone wrong is not necessarily the assertion. In proper cases of selfless assertion, a speaker succeeds in asserting as they ought even if they fail to believe as they ought.
If selfless assertions are proper, then so also are selfless questions. Assertions are selfless when the agent doesn’t believe them (their “doxastic self” isn’t behind the assertion), but the assertion aligns with the speaker’s overall epistemic position anyway. Asked questions are selfless when an agent lacks an IA (or interrogative attitude) in their asked question (their “interrogative self” isn’t behind the asked question), but the question aligns with the speaker’s overall epistemic position anyway.
One reason an agent might not have an IA in a given question is that they don’t actually believe (and so don’t actually know) that the question is sound. Consider the following cases:
Distraught Doctor Redux: Sebastian is in the same situation as before (he can’t get himself to believe that vaccines don’t cause autism, although he appreciates on some level that this is what the evidence overwhelmingly supports). Another doctor in his office used to work with a team of scientists developing vaccines for children. He asks his colleague: How come vaccines don’t cause autism?
Notice that his question presupposes that vaccines don’t cause autism. Every answer to his question has the form, “Vaccines don’t cause autism because….” Since he doesn’t know that vaccines don’t cause autism, he doesn’t know that his question has a true, direct answer. Plausibly, the question he is really wondering isn’t how come vaccines don’t cause autism but whether they do. Even so, his question seems appropriate.Footnote 52
Or consider another case:
Biased Father: John is on a panel judging a poetry competition at a local high school. His daughter Anne submitted one of the poems. John loves Anne’s poem, and even though at a certain level he appreciates that it isn’t the technically most proficient or the artistically most inspired, he reads Anne’s poem with a father’s eyes and can’t rid himself of the thought that it might be the best of the bunch. Even so, he does his best to be impartial when meeting with the other panellists. At the outset, he asks, “So which was the best poem: Judy’s or Jodie’s?”, two clearly excellent poems, thereby passing over the possibility that the best poem was Anne’s.Footnote 53
John − biased as he is by love for his daughter − doesn’t really believe (and so doesn’t really know) that the only plausible candidates for the best poem are Judy’s and Jodie’s. It’s consistent with John’s doxastic commitments that Anne’s poem is the best. Accordingly, it’s natural to think that the question John really has is whether the best poem is Judy’s, Jodie’s, or (rather) Anne’s. Even so, on the assumption that Judy’s and Jodie’s poems are objectively a tier above the rest, and that he is to some degree cognizant of the way his attitudes toward Anne’s poem may be biased, his question seems appropriate to ask, despite the fact that he does not believe (and so does not know) that his asked question is sound.
These cases show that the case for selfless questions is as strong as the case for selfless assertions. Since I’m sympathetic to the permissibility of selfless assertions, I’m also sympathetic to the permissibility of selfless questions. I think, therefore, that the norm I’ve been defending so far should really be modified to allow for selfless questions:
The Position to Know Norm: One ought to: Ask Q only if one is in a position to know that Q has a true, direct answer.
Indeed, a version of this norm has been anticipated by Millson and Risjord (Reference Millson, Risjord, Stovall and Koren2025: 156), who write that “entitlement to undertake an erotetic commitment …appears to be mostly an epistemic affair − i.e., a matter of whether you’re in a position to know that your question has an answer” (although they do not think all askings are ways of undertaking an erotetic commitment).Footnote 54 The position to know is, admittedly, a slippery concept, the meaning of which can vary according to the practical and theoretical needs of a given context. In order to accommodate selfless assertions, the position, or ability, to know, should be understood in such a way that merely psychological obstacles do not always prevent one from being in a position to know something in the sense relevant to the theoretical goals of articulating a norm for asking. Further discussion can be found in Kearl and Willard-Kyle (Reference Kearl and Willard-Kyle2026).
But we needn’t settle the details to see the general point. The important feature of the position to know for our purposes is that one can be in a position to know that p even if one does not actually know (or even believe) it. Sebastian and John are (in some sense) in a position to know that their questions have true, direct answers even though they do not believe so (and so do not know so). The revised version of the norm rightly countenances the possibility of permissible selfless assertions while still affirming that knowledge is central to the speech act of asking.
5. Conclusions
This paper has two goals. First, it has argued for a certain picture of interrogative transmission. Sometimes we ask questions not (merely) to get an answer, but to raise and transmit our question. In this way, asking a question bears some analogy to asserting a proposition. As beliefs in propositions get passed around by testimony, so IAs toward questions get passed around by asking.
Insofar as we can count on people to ask good questions − to ask questions that we may be willing to adopt ourselves on the basis of their so asking − we do so, at least in part, because we take their questions to be good ones. We need some standard or norm for assessing the quality of asked questions, the reliability of conforming to which serves as a barometer of reliability or trustworthiness.
I’ve argued that AKN is too strong to play this role. It’s too strong because it requires us to only ask questions when our addressees know the answer. This fails to account especially for topic-giving questions. Even so, something in the ballpark is right. We should only ask questions that have true answers; indeed, we should only do so when we are in a position to know that our questions are sound.
The evidence for a norm is not just that we typically abide by it. We don’t just comply with the rules: we bend, break, and even flagrantly violate them. We bend the rules to be polite; we flout the rules to make a rhetorical point; we break the rules to deceive. In the right circumstances, these are useful departures from the norm. But the exceptions prove the rule insofar as the usefulness of the exception is parasitic on its distance from the norm. The soundness norm for asking is supported not only by our conformity to it, but by our systematic departures from it.
Acknowledgements
I’m grateful to André Abath, Veronica Campos, Amandine Catala, Erhan Demircioğlu, Arianna Falbo, João Fett, Jason Kawall, Jennifer Lackey, Baron Reed, Mohammed Tayssir Safi, Tim Sundell, and Sarah Wright for asking good and challenging questions about earlier versions of this paper.
Open access
