Doubting Assertion

Abstract

One main argument that has been offered in support of the Knowledge Account of Assertion is that it successfully makes sense of a variety of Moorean-paradoxical claims. David Sosa (2009) has objected to the Knowledge Account by arguing that it does not generalize satisfactorily to make sense of the oddity of iterated conjunctions of the form “p but I don’t know whether I know that p”. Recently, Martin Montminy (2013) has offered a defense of the Knowledge Account. In this paper, I show that both Montminy’s and Sosa’s arguments fail. First, I argue that Montminy does not offer a good reply to Sosa; then I show that Sosa’s objection actually does not constitute a real threat to the Knowledge Account.

One main argument that has been offered in support of the Knowledge Account of Assertion is that it successfully makes sense of a variety of Moorean-paradoxical claims. David Sosa (2009) has objected to the Knowledge Account by arguing that it does not generalize satisfactorily to make sense of the oddity of iterated conjunctions of the form “p but I don’t know whether I know that p”. Recently, Martin Montminy (2013) has offered a defense of the Knowledge Account. In this paper, I show that both Montminy’s and Sosa’s arguments fail. First, I argue that Montminy does not offer a good reply to Sosa; then I show that Sosa’s objection actually does not constitute a real threat to the Knowledge Account.

The knowledge account of assertion (hereafter KAA) claims that an assertion that p is proper only if the speaker knows that p.¹ One main argument that has been offered in support of KAA is that it successfully explains the oddity of Moorean conjunctions of the form

• (1) p but I don’t know that p²

According to KAA, (1) is odd because one can’t assert it without making an improper assertion. In particular, KAA is able to explain the oddity of (1) because it shows that asserting ‘p but I don’t know that p’ involves contradictory conditions. That is, asserting (1) involves the following contradiction: (1) is a proper assertion only if (a) the speaker knows that p and (b) the speaker knows that she doesn’t know that p. Since knowledge is factive, it follows that (b) contradicts (a), and that one can never properly assert a sentence like (1).³

In a recent paper, David Sosa (2009) objects to KAA by arguing that its explanation of the Moore’s Paradox “does not generalize satisfactorily” (Sosa 2009, 270). That is, he claims that KAA fails to explain the oddity of Moorean iterated conjunctions of the form:

• (2) p but I don’t know whether I know that p.

Sosa’s objection to KAA is that the “adequacy” of KAA’s account of the oddity of (1) “is challenged by the oddity of” (2). In particular, Sosa maintains that “[w]e do not adequately understand what is wrong with” (1) “if our explanation does not account for what is wrong with” (2) (Sosa 2009, 270). Since asserting “I don’t know whether I know p” is not inconsistent with one’s knowing that p, KAA cannot generalize to explain the oddity of assertions of (2) without appealing to the controversial “KK principle” (when you know that p you also know that you know that p). This is a problem for KAA because most recent proponents of KAA reject the KK principle. ⁴

Sosa maintains that since KAA does not generalize to make sense of the oddity of (2), the sort of argument that proceeds from the Moore’s Paradox threatens to generate a regress: the knowledge account of assertion can only explain the oddity of (1), so in order to explain the oddity of (2) we need to refer to the knowledge-that-you-know account of assertion, and so on and so forth. In Sosa’s own words, “the knowledge account will have to be strengthened implausibly, to require, for every level, knowledge that you know… that you know that p” (Sosa 2009, 271). Sosa concludes that the argument in support of KAA that proceeds from the Moore’s Paradox fails.

In a recent paper, Martin Montminy (2013) argues that proponents of KAA have a way to explain the oddity of (2) without subscribing to the KK principle.⁵ What is wrong with (2) is that the speaker does not know whether her assertion is appropriate and cannot properly defend her assertion.⁶ In this paper I show that, even though Montminy’s solution fails to address Sosa’s worry, KAA can still be rescued. In the first three sections of the paper, I present three independent objections to Montminy’s argument and after that I offer my own solution.

Epistemically Improper Assertion

Martin Montminy (2013) has offered a solution to the oddity of (2) that derives from KAA. Assuming KAA, a speaker who asserts (2) “is in no position to know that her assertion is appropriate”. Thus, (2) displays an “epistemic deficiency” of the speaker and, as a result, the speaker can’t appropriately defend her assertion. This is because a speaker who asserts (2) can’t appropriately respond “yes, I know that p” when challenged by the questions “Do you know that p?” (Montminy 2013, 829). On this view, the explanation of the oddity of (1) stems directly from KAA, whereas the explanation for the oddity of (2) derives from KAA (although less straightforwardly). The validity of the argument in favour of KAA that comes from the Moore’s Paradox is thus preserved.

Unfortunately, I believe that Montminy’s solution faces a dilemma: either it fails to address Sosa’s real concerns or it is superfluous, and this whole issue has been poorly framed from the start. To see what I mean, first recall that an assertion can be appropriate or inappropriate in ways (e.g. morally, prudentially) that go beyond epistemic impropriety. To illustrate, telling my boss he is bald may be an epistemically appropriate assertion, but it definitively fails a norm of prudence. And so on.

What’s more, it is worth paying attention to the distinction between assertions that somehow fail because of the speaker’s epistemic position from assertions that are epistemically improper. It is important to draw this distinction if we want to make sense of the following distinct phenomena⁷:

• a. a speaker’s epistemic position lacks some epistemic features (e.g. knowledge, justification), and this makes it the case that her assertion is improper in some, non-epistemic sense.

• b. a speaker’s epistemic position lacks some epistemic features (i.e. knowledge), and this makes it the case that her assertion is epistemically improper.

I will illustrate my point here by means of the following example:

AFFAIR⁸

Husband: Why didn’t you say she was having an affair? You’ve known for weeks.

Friend: Well, since I knew the damage it would cause to your marriage, it wouldn’t have been right for me to say anything before I was absolutely sure.

Although Friend admits that he knew about the affair, he claims that it would have been inappropriate to let Husband know given his epistemic position (i.e. knowledge but not certainty). Now, it is possible to argue, and this is the key point here, that the notion of ‘propriety’ at play here is not epistemic. It is true that Friend’s epistemic position was “suboptimal” because he lacked certainty (Coffman 2010, 13). But it was suboptimal only with respect to some normative considerations that, I contend, may have nothing to do with epistemology. In AFFAIR these are most probably moral considerations: telling Husband that his wife was cheating on him without being absolutely sure would be equal to making an assertion that is – arguably – morally improper. The proposal then is that, in AFFAIR, the reason why Friend did not tell Husband about his wife’s affair was some kind of moral concern related to his epistemic position.

Thus, from the fact that a speaker’s epistemic position intuitively harbors some epistemic shortcoming it does not follow that that assertion is also epistemically improper. That is, it may well be that in AFFAIR, the speaker’s epistemic position is strong enough to make Friend’s assertion epistemically proper. The suggestion here is that in this case his assertion would have been improper in some non-epistemic sense and because of his epistemic position.

This distinction is relevant to our topic here. If we look back at our dubious conjunction (2), Montminy’s solution indicates that asserting (2) is odd because it shows that the speaker bears some defective epistemic relation to the content of her claim, i.e. she does not know whether she knows it. However, from that it does not automatically follow that her assertion’s failing is epistemic (as it does not follow in AFFAIR).

So, one may ask, in what sense is (2) inappropriate? What kind of inappropriateness results from the speaker’s defective epistemic position? I think there are two answers Montminy can offer to this. The first one is to say that, although asserting (2) is not epistemically improper in some primary sense, there is still something epistemically improper in making that assertion. Let me explain. Theorists have individuated at least two different levels of propriety/impropriety for norms: primary and secondary propriety/impropriety. If an act A is governed by some norm N, primary propriety/impropriety is determined by whether A actually abides to N; instead, secondary propriety/impropriety is determined by whether the agent doing A reasonably believes that A conforms to N (DeRose 2009).⁹ When applied to assertion, primary impropriety takes place whenever there is a violation of the primary norm of assertion. This is the norm that established when an assertion is proper or improper qua assertion. But from a primary norm we can derive a secondary norm, according to which one should not make an assertion unless one has a reasonable belief that one’s assertion respects the primary norm. This is sometimes explained with the idea that a speaker is blameworthy if she does not reasonably believe her assertion is proper in the primary sense.¹⁰ Now, Montminy may say that is that (2)’s epistemic oddity is not the result of a primary or secondary impropriety,¹¹ but that there is a further sense of epistemic infelicity that occurs when one does not know that one’s assertion is primary proper. And this explains the oddity of (2) and the epistemic violation that generates such an oddity. The problem with this solution, however, is that it falls prey to Sosa’s worry that, if we go down that path, then we will need a fourth, fifth, sixth, and so on level of epistemic infelicity each time we reiterate the second conjunct of (2) and form higher order levels of knowledge (Sosa 2009, 271).

The second route is to focus on Montminy’s claim that if one does not know that her assertion is appropriate one can’t appropriately defend it. Thus, (2) violates the following norm:

• (C) one should appropriately defend one’s claim when (appropriately) challenged.

Is (C) an epistemic norm? I don’t have a clear-cut way to single out epistemic improprieties from other kinds of improprieties. But it is at least dubious that the inability to defend one’s claim makes one’s assertion epistemically improper, for two reasons. First, note that (C) is a, so to speak, ‘performance norm’, i.e. a norm that requires us to do something. As result, it can be violated in various ways, some of which have very little to do with one’s epistemic position. To illustrate: a speaker can violate (C) by intentionally refusing to give reasons for her claim, even if she in fact has those reasons. Her epistemic position is thus on par, but she still violates (C). So violations of (C) are thus not necessarily epistemic violations.¹² In contrast an epistemic norm, like for instance the knowledge norm of assertion promoted by KAA, is a norm exclusively concerned with one’s epistemic position and can be violated only by failing to know the content of one’s claims. What’s more, (C)’s job seems to regulate our conversational practice and mutual understanding among speakers. We expect speakers not only to be in good epistemic standing vis-à-vis the claims they make, but also to show that they are so positioned; in this way they can defend their credibility as speakers while also putting others in a position to be persuaded by those ideas and claims. Thus, (C) seems to be better interpreted as a norm that allows communication among speakers and thus as a conversational norm.

So why is uttering (2) odd then? A speaker who asserts (2) makes it explicit that she does not know whether her assertion that p is appropriate (assuming KAA is right, of course). On Montminy’s reading, that amounts to saying one is not in a position to defend one’s claim. Thus, in uttering (2) the speaker asserts that p, and then reveals that her epistemic position with respect to her assertion that p. This generates a conversational impropriety according to (C). And this may be the explanation why (2) sounds odd.

Unfortunately, this solution raises the following dilemma: either his explanation of the oddity of (2) is superfluous as a defense of KAA or is unable to address Sosa’s worry. Let me explain. It seems that one should interpret Sosa as arguing that the oddity of (2) reveals that there is something epistemically improper with (2). Thus Montminy’s solution would be fruitful only given the assumption that (C) is an epistemic norm and violating (C) is committing an epistemic impropriety rather than a conversational one. However, as I mentioned above, we have no support for this assumption. Thus Montminy’s solution does not seem able to address Sosa’s worry: it simply does not show that (2) is epistemically improper based on KAA.

Alternatively, one could see the argument offered by Sosa as simply requiring that the oddity of (2) reveals that (2) is in some sense improper. As explained above, Montminy’s solution may be able to account for why (2) is in some sense improper. However, if that were Sosa’s point, then it unclear it would constitute an objection to KAA, as the “adequacy” of KAA’s account of the oddity of (1) would not be challenged by the oddity of (2). (Sosa 2009, 270) KAA is only committed to saying that lack of knowledge is what makes an assertion an epistemically improper assertion. That is, KAA does not aim to offer an account of assertion that makes sense of any kind of impropriety related to the speaker’s epistemic position. KAA talks only about epistemic impropriety. So in this case I doubt Sosa’s objection would not represent a worry for KAA and no explanation of the oddity (2) that appeals to KAA would be needed.¹³ Sosa might reply to this by saying that the infelicity involved in dubious assertions like (2) is, even if not epistemic in nature, similar enough to the infelicity involved in (1) that our strategies for explaining the two sorts of infelicity should be relatively unified. Assuming this is true and that we can agree on what ‘similar enough’ amounts to here, I believe that this reply still puts pressure on Montminy’s solution. Indeed, it seems clear that, even if Montminy’s solution stems from KAA, based on my reconstruction it itself presupposes that the two infelicities are not at all similar. First, as pointed out at the beginning KAA solution for (1) points to a contradiction as the source of the oddity of (1), but for Montminy there is no contradiction going on in (2). Second, for Montminy the oddity of (2) has nothing remotely to do with (2) being appropriate or not or with worries that it may not be appropriate, again contrary to what happens in (1).¹⁴ So if (1) and (2) really harbor similar problems, his solution cannot make sense of that.

Defending Assertion

In the following two sections of the paper I will look at Montminy’s diagnosis of the oddity of (2) and argue that it is quite problematic. My goal is to show that KAA cannot and should not adopt Montminy’s view. This analysis will bring me to formulate my own solution to the puzzle.

One argument in support of KAA points to the fact that “Do you know that?” and “How do you know that?” are appropriate ways to challenge an assertion. And as Montminy rightly observes, “[a] speaker who asserts (2) can’t properly answer ‘yes’ to this challenge, since this would entail that she knows that she knows that p.” (Montminy 2013, 829) From that Montminy concludes that a speaker who asserts (2) reveals that she can’t defend her claim, and that’s why (2) sounds odd. And he is not alone in thinking that. For instance, in a recent paper Greco (2014, 667 fn.) has argued that in cases of dubious assertions, “while [the speakers] will be able to permissibly assert that p, if their permission to assert that p is challenged, they will not be able to permissibly defend themselves.” Thus, Greco seems to believe that dubious assertions are problematic because they show that the speaker lacks knowledge of whether her assertion is appropriate.

But this is too strong. Even if a speaker does not know whether her assertion is appropriate, that does not necessarily affect her ability to respond to challenges and permissibly defend her claim. First, it is unclear that a speaker can adequately respond to “Do you know?”-challenges only if she can flat out assert that her assertion is appropriate. Let’s consider our conversational practice. It is true that such “Do you know?”-challenges are common (as Williamson and others have pointed out). In fact, when our assertions are challenged by those questions, there are many cases in which we feel very confident and we reply to those challenges by saying “yes, I know that” or “of course, I do”. However, we sometime don’t feel sure that we know. Still, it seems we can appropriately respond to “Do you know?”-challenges by uttering hedge assertions such as “I think so”, “I believe that I do”, “I would be surprise if I didn’t”.¹⁵ We generally accept those as appropriate replies to challenges, even if the speaker doesn’t flat out assert that she knows. So the speaker’s inability to flat out assert that she knows cannot explain the oddity of her uttering (2). ¹⁶

There is a further point. Montminy maintains that in order to appropriately defend her claim, the speaker should necessarily be able to respond to “Do you know?”-challenges. In contrast, I believe we need to keep distinct the speaker’s ability to respond to those challenges from the issue of whether or not she is able to appropriately defend her claim. The latter issue, I contend, has to do with the speaker’s ability to offer reasons to justify what she said¹⁷ and show that it is likely that she possesses the required knowledge. In contrast, at least on Montminy’s view, the former point depends on whether the speaker is able to flat out assert that she knows that the content of her assertion is true. Note that the two things are not necessarily correlated. For instance, take the case in which, given her evidence, the probability that a speaker knows that p is very high, while the risk of her making an improper assertion is quite low. And she knows that. If such a speaker makes the assertion that p, she is able to defend it by offering evidence and explaining why she thinks that she knows that p. That’s true even if she is not in a position to respond to challenges by claiming that she knows that p. A defense of the assertion that p may count as epistemically appropriate even if the speaker can only offer evidence for p but is unable to flat out say that she knows that p.

To make this more vivid let’s consider the following scenario¹⁸: we are on a trip in some remote land and we start a conversation about soccer. I make the claim that Germany did not win the World Cup in 1982. Indeed, I believe (and also know) that Germany did not win the World Cup in 1982. However, you disagree and challenge my claim by asking, “Do you know that?”. Even though you are not a soccer expert, your challenge makes me lose some of my confidence. Therefore, when you ask me whether I know that Germany did not win the World Cup in 1982, I can’t flat out say ‘yes’. However, that does not mean I can’t appropriately defend my initial claim. I can still defend my assertion by mentioning the fact that I am very much into soccer, that I have watched innumerable soccer games and that I am a very reliable source when it comes to the sport. This counts as a proper defense of my claim, unless you are able to offer me some valid reasons to think that my assertion is improper and that I don’t possess the relevant knowledge.

So, I conclude, a speaker who lacks knowledge that her assertion is appropriate is unable to fully respond to some challenges by flat out asserting that she knows. That’s true. However, this does not mean she is unable to defend her claim: it seems that there is a lot she can do to support it anyway. And it is unlikely that the speaker’s inability to fully address conversational challenges is enough to explain the oddity of asserting sentences like (2).

Assertion as Speech Act

I now will argue that supporters of KAA should not accept Montminy’s diagnosis of the oddity of (2) because it transforms KAA in an implausible theory about speech-acts. This is especially problematic if we reflect on the fact that Montminy’s proposal is actually meant to defend KAA from a pressing objection against it.

According to KAA knowing that one knows that p is required to be able to properly assert that one knows that p. And on Montminy’s view being able to assert that one knows that p is necessary to appropriately defend one’s claim that p. So it seems that KAA can explain what is wrong with asserting (2): the speaker “may properly assert (2) but she cannot properly defend” her claim that p (Montminy 2013, 829). Unfortunately, if true, this would make KAA an implausible theory of assertion that requires that, in order to appropriately defend her assertion that p, the speaker has to make a stronger speech act with the same content.

Let me explain. Austin (1946) famously argued that going beyond the simple claim “p” to the assertion “I know that p” we take a “further plunge”. Austin explained that if I say that I know that the meeting starts at four, then if I turn out to be wrong, I am “liable to be rounded on by others” more harshly than if I had simply said that the meeting starts at four (Austin 1946, 172). Austin took “I know that p” to be a performative as he argued that there is a relation between “I know” and “I promise”.¹⁹ His view was that if you utter the words “I promise to A”, then you perform an act, i.e. you promise to A. Similarly, to say “I know that p” is to do something, that is, to “give others my word” that p is true, thereby authorizing them to repeat “p” (Austin 1946, 171).

I think Austin’s view is very intuitive here as it makes sense of the fact that in communication we interchangeably use expressions such as ‘I know that’, ‘I guarantee you that’, ‘I promise you that’. Also, when we say ‘I know that is true’, we seem to be doing more than just asserting that it is true. We are really putting our full authority behind it. ^{20 21}

So requiring that a speaker be in a position to say “I know that p” in order to appropriately defend her assertion that p is tantamount to saying that she can appropriately defend her assertion that p only if she is able to make a stronger speech act with the same content. But this raises various problems.

One: If, to defend an assertion that p, I need to be in a position to make a stronger move with respect to p, why would I not be making that stronger move in the first place? Indeed, one may even argue that I am actually somehow deceiving my audience if I make a weaker speech act when I am actually prepared to make a stronger one with the same content.

Two: Montminy’s idea threatens to lead KAA straight into a regress: if I can defend my assertion only by making a stronger speech act, then I can defend my stronger speech act only by making a possibly even stronger speech act. But then that speech act can be challenged as well. So I need to be prepared to make a speech act that is possibly even stronger than that. And so on. That means that, when I assert that p, I need to be prepared to make an indefinite number of (possibly stronger) speech acts to defend my initial claim. And this sounds quite implausible.

Three: assertion is an alethic speech act and in many respects is similar to other alethic speech acts such as conjecturing, guessing, guaranteeing, and the like.²² On Montminy’s view, KAA sets assertion apart from other speech acts in a way that is, I believe, problematic. Consider the speech act of conjecture. According to a prominent view on speech acts, conjecturing is a less demanding speech act than asserting. That is, asserting “extracts more credibility” than conjecturing because assertion is governed by tougher epistemic standards. (Turri 2010, 92) This can be explained by pointing to the fact that asserting that p requires that you know that p is true (assuming KAA is correct), whereas conjecturing requires that you have at least some evidence that p is true. Accordingly, it is legitimate to conjecture that p without knowing that p, i.e. without being in a position to assert that p. In order to properly make a conjecture that p I need to have at least some evidence that p is true, but I don’t need to know that p is true. And what’s more, in order to properly defend my conjecture I don’t need to be in a position to take a further plunge and make a stronger speech act with respect to the proposition “p”. Let’s see why. Consider the following scenario:

TEST:

Mike is a very good student who usually performs very well on math tests. After taking a particularly difficult math test, Mike tells his friend that since he is quite good at math and was well-prepared for the test, he thinks he passed it. However, he doesn’t know for sure and so he will need to wait till the grades come out.

In TEST Mike is conjecturing that he passed the exam while also admitting that he does not know whether he passed it. He prepared well for it and he usually does well on those kinds of tests, so his conjecture seems to be perfectly legitimate. If challenged he can defend himself by offering the evidence he has for making that conjecture. Anybody will agree Mike does not need to be in a position to assert that he passed the exam in order to properly defend his conjecture. That is, in order to defend his conjecture he is not required to be in a position to take a further plunge and make a stronger speech act than conjecturing. Indeed, there is no norm that demands that he undertakes a stronger commitment than the one his conjecture actually requires.

Now, given the similarity of assertion with other speech acts, this must be the case for assertion as well. However, this is not what Montminy’s view predicts for assertion. That is, according to Montminy, KAA is committed to saying that if a speaker can’t respond to challenges by asserting that she knows that p, then she can’t properly defend her assertion that p. However, if we follow Austin, we are committed to saying that uttering the words “I know that p” is making a stronger speech act than just asserting that p: it is promising that p is true. On Montminy’s view, in order to properly defend an assertion that p, a speaker needs to make the promise that p, and thus, on Montminy’s view, KAA makes assertion work differently from other alethic speech acts.

Now, the problem with this emerges clearly once we notice that there is actually a compelling reason for why the speech acts work that way. One feature of alethic speech acts is that they extract different levels of credibility in relation to the epistemic norms that govern them. Speakers can be blamed and lose credibility in their community if they violate such epistemic norms. So to avoid losing credibility speakers have the option to make speech acts that extract a lower level of credibility instead of making more demanding speech acts (e.g. guarantee, swear). That is, the system is such that a speaker has the option to go for something less committal in order to preserve her credibility as a speaker. In contrast, this system would become seriously dysfunctional if speakers were forced to make more demanding speech acts just to defend their initial – less demanding – speech act. This is because it would mean that making a weak speech act does not allow the speaker to preserve her credibility. And thus this graded system of speech acts would lose its rationale.

Assertion is an alethic speech act and is part of this system of speech acts. So it would be surprising if it turned out assertion did not work as the other alethic speech acts do. And indeed, this would make assertion a dysfunctional speech act that extracts the same level of credibility of stronger speech acts and entails the same risk to incur in blame and criticism. But this is what KAA would end up saying, if it accepted Montminy’s proposal. This is why I believe KAA should in fact reject it.

A Reply to Sosa

So far I have argued that Montminy’s solution should not be adopted by KAA. Does this mean that, since KAA lacks any compelling reply to Sosa, Sosa’s objection is on target? I believe the answer is no: Sosa’s objection fails just as much as Montminy’s reply to it. This is because there is no reason to expect that KAA should provide a solution to dubious assertions like (2), so no explanation of the oddity (2) that appeals to KAA is needed.

Let’s backtrack for a moment. We have two claims, (1) and (2), that both sound odd somehow. That I won’t deny. Also, I assume that, if the oddities of (1) and (2) are in the same ballpark and points to some similar underlying problem, then a view that wants to offer a plausible explanation for (1) should also be able to address (2), in the sense that its strategies for explaining the two sorts of oddities should at least be relatively unified. Unfortunately, I don’t see any reason to believe that the two oddities are in fact similar enough to require such a unified explanation.

Let me explain what I mean. Admittedly, Sosa has told us very little about why we should take the oddities of (1) and (2) to be similar. In contrast, there is indication that (1) and (2) are different in some important ways. First, the structure of the two iterated conjuncts (1) and (2) is very different and the oddity of (2) does not seem to seem to produce a clash between the two conjuncts as (1) does. As pointed out by Matthew Benton (2013) in a recent paper, whereas the construct of (1) has the form: ‘p but I don’t know that p’, (2) shows a different form, namely ‘p but I don’t know whether [that] q’. In addition, assertions of Moorean-type sentences such as (1) produce clash, whereas (2) seems to be just really odd and clunky thing to say, but no clash between the two conjuncts is produced (DeRose 2009, 208).

Second, it seems that, if one offers the appropriate context, the oddity of (2) can be explained away without producing a hedged assertion. If we consider for a moment the soccer example mentioned above, we see that it may even be possible to make the felicitous assertion, “Germany did not win the World Cup in 1982” followed up by the admission that one doesn’t know whether one knows while also offering reasons to think that nevertheless it’s likely that one knows. I argued that offering those reasons is a way of defending one’s claim without giving up on it. Alternatively, it is possible to explain away the oddity of (2) and thus see it as a felicitous assertion by clarifying its significance. As I pointed out above, in claiming that one knows that p, one in general stakes more credibility than in merely claiming that p. Following Austin (and others) we say that saying “I know that p” is tantamount to guaranteeing that p. If this is right, then we can actually explain away the oddity of uttering a proposition like (2) simply by pointing out that we need to understand those who utter (2) as making the assertion that p and adding that they are not prepared to also guarantee that p. In other words, in (2) the speaker asserts that p while also wanting to clarify that she is not guaranteeing that p. All this points to a clear disanalogy between (1) and (2): whereas asserting (2) can be rendered felicitous by introducing some explanations and the right context, those explanations and contexts do not make asserting (1) felicitous. So the two assertions are odd in importantly different ways. ²³

And finally, let me point out that the very fact that KAA seems to have no good explanation for the oddity of dubious assertions like (2) is itself evidence that (1) and (2) must differ in some important way. First, note KAA can explain why an assertion of (1) is inappropriate. As I said, Sosa’s objection trades on the assumption that the oddity of (1) and (2) are the same or at least similar enough. But that KAA cannot explain the oddity of (2) does not automatically disconfirm it. This is because Sosa has offered us no clear reason to accept his assumption and thus KAA’s ability to explain the oddity of (1) not only provides support for KAA but also offers a reason to believe that there is an infelicity that (1) has and that (2) does not have. This point can be made even stronger once we notice that one of the reasons the KAA has enjoyed so much popularity lately is that it is able to make sense of so many of our linguistic intuitions about assertion. These are well-known in the literature but let me mention few of them. Beyond traditional Moorean sentences KAA can make sense of important conversational challenges. As pointed out above, when a speaker makes an assertion, she is thereby open to challenges, such as “How do you know that?”/ “Do you know that?”, which test her authority to make the assertion in the first place. Also, assertions may be rejected by saying “You don’t know that!”. Such challenges and criticisms presuppose and make sense only if the speaker is required to know what she asserts, and thus they strongly support KAA (Williamson 2000, 252). What’s more, when someone asks us whether p is true, it is usually perfectly appropriate to respond by saying “I don’t know”. This makes perfect sense according to KAA: by doing that we are informing this person that we lack the necessary authority to answer her question by making an assertion (Reynolds 2002, 140). Finally, if someone asserts that p, and then it turns out that she did not know that p, we feel perfectly entitled to resent that and accuse her of having spoken without the necessary authority (Unger 1975, 260–3). Again, KAA can make perfect sense of why we feel that way.²⁴

Given KAA’s strong explanatory power, we have reason to believe that contra Sosa what goes on with dubious assertion such as (2) is in fact very different from what goes on with more traditional Moorean sentences. After all, we are producing an inference to the best explanation here, and KAA is able to explain a lot of the data. Hence, at this point we have no compelling reason to expect that KAA should provide a solution to dubious assertions like (2), and thus Sosa’s objection is no serious threat to that view.²⁵