No Infelicity for the Infallibilist

: According to probability 1 infallibilism (henceforth, infallibilism ), if one knows that p, then the probability of p given one ’ s evidence is 1. Jessica Brown (2013, Analysis, 73 , 626 – 635; Fallibilism: Evidence and Knowledge , 2018) has recently argued that infallibilism leads to scepticism unless the infallibilist also endorses the claim that if one knows that p, then p is part of one ’ s evidence for p. By doing that, however, the infalliblist has to explain why it is infelicitous to cite p as evidence for itself. And yet, the infallibilist does not seem to have a satisfying explanation available. Call this the infelicity challenge for probability 1 infallibilism . By exploiting the distinction between the justifying and the motivating role of evidence, in this paper I argue that contrary to ﬁ rst appearances, the infelicity challenge does not arise for probability 1 infallibilism. However, after anticipating and resisting two objections to my argument, I show that we can identify a different version of infallibilism which seems to face a problem that is even more serious than the infelicity challenge. Probability infallibilism If S knows that p, then the probability of p given S ’ s total evidence


Infallibilism and the Infelicity Challenge
INFALLIBILISM about knowledge has a bad reputation for it is often thought to easily lead to scepticism. In a nutshell, the worry seems to be that infallibilism sets standards for knowledge that are too high to be realistically achieved. There are many ways to define infallibilism. However, given its relevance in the contemporary debate, for the purpose of this paper, I will focus my attention on the following version of infallibilism, called probability 1 infallibilism 1 : spirit of infallibilism, 3 especially when combined with an externalist view about the nature of evidence that attracted a lot of attention over the past two decades, which she calls the sufficiency of knowledge for evidence view (Brown, 2018, ch. 2): Sufficiency of knowledge for evidence (SKE): if S knows that p, then p is part of S's evidence. 4 In fact, as Brown rightly points out, SKE entails P1I. For if every proposition p that one knows goes into one's "box of evidence," then the probability of p given one's total evidence will be trivially one insofar as one's total evidence will include p itself. Given the popularity of SKE, if probability 1 infallibilism faces a sceptical challenge, then it is worth taking it seriously. 5 Crucially, according to Brown, given that the evidence that underwrites our propositional knowledge generally does not entail the target known proposition, the only way for the infallibilist 6 to avoid the sceptical threat is to endorse the following thesis concerning evidential-support: Sufficiency of knowledge for self-support (SKSS): If S knows that p, then p is part of S's evidence for p.
And yet, if the infallibilist wants to successfully resist the sceptical worry by endorsing SKSS, then she will have to provide an explanation of what Brown calls the infelicity data, namely, the fact that it is infelicitous to cite p as evidence for itself (Brown, 2018, 45-7). I call this the infelicity challenge.
As I will explain later, Brown anticipates two potential strategies to meet the challenge. The infallibilist could either question the infelicity data (strategy 1) or accept the infelicity data while accounting for its plausibility in a way that is compatible with the literal truth of citing p as evidence for itself (strategy 2). To defend strategy 1, the infallibilist might appeal to the notion of self-evident propositions. To defend strategy 2, she could appeal to a pragmatic explanation. Unfortunately, Brown argues that none of these two strategies is successful and SKSS should thus be rejected: 3 One might think, against Brown, that probability 1 infallibilism does not really capture the spirit of infallibilist theories of knowledge. I think this is a very reasonable worry to have. However, for the purpose of this paper, I will grant Brown that probability 1 infallibilism is in fact the best formulation an infallibilist can come up with. 4 Many epistemologists endorse SKE by defending the idea that p is part of one's evidence if and only if one knows that p (E = K). See Williamson, 2000, Hyman, 2006, Littlejohn, 2008, Logins, 2017, Fratantonio, 2018 Another reason to take this challenge seriously is that, contrary to internalist accounts of evidence, externalist theories like SKE are generally thought of as being the antidote to scepticism (Cf. Williamson 2000, ch. 8). 6 From now on, when I refer to "infallibilism," I am always referring to probability 1 infallibilism.
[T]here is a range of evidence which provides some reason to doubt the SKSS. First, it's ordinarily infelicitous to cite a proposition as evidence for itself, even if it's known. Second, in a variety of debates, such as "forgotten evidence" debate, philosophers don't seem to even consider the possibility that every proposition which is evidence is evidence for itself. (2018,53) This paper is threefold. The main aim of this paper is to argue that the bad reputation probability 1 infallibilism has gained is undeserved: it does not lead to scepticism and it is not threatened by the infelicity challenge. To do soand this is the second aim -I will argue that the infelicity challenge rests on an ambiguity of the evidence-for relation underpinning SKSS (section 2). In turn, having disambiguated between two different interpretations of the evidence-for relation will allow me to show that there are independent reasons why the foregoing strategies are not successful (section 3). Third, before concluding the paper (in section 5), I will briefly anticipate and resist two objections to my argument. In addressing and resisting the first objection, I will show that, by drawing on the two notions of evidence-for relation distinguished in section 2, we can identify a different formulation of infallibilism, which, contrary to probability 1 infallibilism, seems to be threatened by a problem which looks worse than the infelicity challenge (section 4). In addressing and resisting the second objection, I will show that, contrary to first appearances, the infallibilist endorsing probability 1 infallibilism is not committed to any of the formulations of SKSS defined in section 2.

No Infelicity for Probability 1 Infallibilism
Following a distinction popular amongst philosophers of action as well as epistemologists, we can separate a justifying from a motivating role of evidence (Cf Alvarez, 2010). On the one hand, one's evidence plays a justifying role insofar as it is what makes a target proposition justified. On the other hand, evidence plays a motivating role insofar as it is that evidence in the light of which one believes a target proposition. 7 Given these two roles evidence can play, we can disambiguate between the following: Evidence-for M : a proposition e is motivating evidence for p for S iff e is the evidence in the light of which S believes that p. 8 Evidence-for J : a proposition e is justifying evidence for p for S iff e stands in an evidential-support relation towards p so as to make p justified.
We can appreciate the distinction between evidence-for M and evidence-for J by considering how they are intertwined with the (evidentialist) notions of propositional and doxastic justification. 9 Roughly put, propositional justification is a property of a proposition given one's evidence, such that a proposition p is justified for one iff one's evidence stands in an appropriate evidential-support relation towards p (e.g., e makes p sufficiently likely). 10 An orthodox way of understanding doxastic justification in terms of evidence, instead, is to take doxastic justification to be a property of a belief such that one's belief that p is justified only if one's belief that p is appropriately based on one's supporting evidence. It is in virtue of the justifying role of evidence that a proposition can be propositionally justified for one. It is in virtue of both the justifying and motivating role that evidence plays that one's belief that p can be doxastically justified.
With this distinction in play, we can ask: (i) What notion of evidence-for underpins the infelicity data? (ii) What notion of evidence-for is entailed by infallibilism?
To answer the first question, we first need to clarify what kind of infelicity is involved in the infelicity data. To do so, let us consider Brown's own examples:

Detective Morse
[S]uppose that in the course of his investigation into the recent burglary at the Central Jewellery Store, the detective Morse comes to know by eyewitness testimony that the notorious burglar, Burglar Bill, was in the vicinity of the Central Jewellery Store just before the theft. Suppose that we ask Morse, "What evidence do you have for the claim that Burglar Bill was in the vicinity of the Central Jewellery Store just before the theft?" In reply, it seems inappropriate for Morse to say, "Burglar Bill was in the vicinity of the Central Jewellery Store just before the theft. […][I]t's in general infelicitous to reply to a request for evidence for p by citing p, even if p is part of one's evidence." (2018, pp. 40, 50,51).
Risk 11 Suppose that, in conversation, a friend asserts that I should cut down on alcohol since alcohol is a cancer risk.
[…] I ask my friend for evidence for the claim, i, that alcohol is a cancer risk. If she it follows that if one uses a false proposition q as premise in deliberation, then q will not count as motivating evidence. These complications, however, will not matter for the purpose of this paper. 9 For a classic defence of evidentialism see Conee and Feldman (2004). For a classic alternative to evidentialism about justification see Goldman's process reliabilism (Goldman, 1979). 10 Alternatively, one could define propositional justification as follows: if S were to believe that p, then S's belief that p would be justified given one's total evidence iff one's evidence appropriately supports p (regardless of whether one in fact believes that p). Thanks to an anonymous referee for pointing this out to me. 11 Labels are mine.
replies merely by citing i, then she is implicitly claiming that the proposition that alcohol is a cancer risk is evidence for the proposition that alcohol is a cancer risk. But this claim doesn't seem to be true, but rather false. (Indeed, it is hard even to hear her reply as claiming that the proposition that alcohol is a cancer risk is evidence for that very proposition. Instead, it simply sounds like a reiteration of i which, in the context, constitutes a refusal to provide any evidence whatsoever for i.). (2018, pp. 55-56; 2013, p. 629) 12 The foregoing cases show that it is that distinct evidence e ≠ p that the inquirer is interested in when asking for evidence. After all, Brown says, in these kinds of contexts, "it's surely clear that I am asking for evidence for the relevant claim [p] rather than [p] itself" (2013, p. 629). The infelicity data as Brown presents it thus concerns what would constitute a felicitous reply to an evidence-request of the form: "what is S's evidence for her claim that p?". A felicitous answer to this question will be one that gives us an insight about the evidence S took as favouring believing p over not-believing p. For instance, as Joyce (2004)  Given what said above, the infelicity challenge would be a real challenge for the infallibilist only if she had to endorse SKSS understood with evidence-for M : SKSS-M: if S knows that p, then S believes that p on the basis of p.
Crucially, probability 1 infallibilism remains silent on what kind of evidence one should base one's belief on in order to gain knowledge. However, it is plausible to take probability 1 infallibilism as resting on a probabilistic account of evidential support, for example, one on which e offers evidential support to p iff e makes p likely over a certain threshold. 13 A plausible interpretation of probability 1 infallibilism is thus one that takes it to be the claim that knowledge of a proposition p requires that p offers adequate evidential support to p. In other words, it is plausible to take infallibilism to be committed to a justifying reading of SKSS: SKSS-J: If S knows that p, then p offers evidential support to p. 14 But if this is so, then the infelicity challenge as Brown presents it does not arise for infallibilism. By moving from a claim about the evidential-support relation (what P1I and SKSS-J are concerned with) to a claim about what is proper to cite as evidence for a proposition p to an evidence-request (what the infelicity challenge and SKSS-M are concerned with), Brown seems to be moving from a claim about the justifying role of evidence to its motivating role. The problem with Brown's infelicity challenge is that it takes something about a probabilistic relation between one's evidence and a target proposition to licence a claim about what is proper or improper to cite as evidence for a proposition as a response to a request for evidence. Crucially, whereas infallibilism can be plausibly seen as being committed to something about the former (e.g., because it can be understood in the light of SKSS-J), it is silent on the latter issue. But then why should the infallibilist be expected to provide an explanation of the infelicity data? The infallibilist could grant that it is in fact infelicitous to cite p as evidence for itself in the context of evidence-request without having to come up with any explanation of why this is the case given that this is a phenomenon that is not entailed by her view.
To sum up: A plausible interpretation of infallibilism is in terms of evidencefor J , and yet it is evidence-for M that underpins the cases offered by Brown that give rise to the infelicity data. On this diagnosis, the infelicity challenge rests on an ambiguity of the notion of evidence-for.

Self-Evidence and Pragmatics
In this section, I appeal to the distinction between evidence-for M and evidencefor J to show that strategy 1 and strategy 2 are not successful for reasons that are independent of the plausibility of SKSS. This will help me to further corroborate my diagnosis of where Brown's infelicity challenge goes wrong.
Strategy 1: According to the first strategy, perhaps the infallibilist could just deny that it is always infelicitous to cite p as evidence for itself. To support this claim, the infallibilist might appeal to the existence of a special class of propositions that are generally taken to be evidence for themselves, or self-evident, for example, self-verifying propositions and propositions about one's experiences.
14 In fact, I will later show that, as defined, P1I is compatible with the rejection of SKSS-J.
Brown thinks there are good reasons to doubt the efficacy of this move. She argues that if self-evident propositions were a genuine possibility, then it is puzzling why this option is not even considered by internalists to avoid the notorious problem of forgotten evidence: Forgotten evidence S acquires a justified true belief that p at time t1 on the basis of some supporting evidence e. At a later time t2, S has forgotten her original evidence for p, but she retains her belief that p.
The problem that forgotten evidence is supposed to raise is that, although S has forgotten her original evidence for p, we want to say that she is still justified in believing that p at a later time t2. As Brown rightly points out, cases of forgotten evidence have traditionally been raised as a counterexample to internalistevidentialists views on which one's justification supervenes on the evidence one has (e.g., Conee & Feldman, 2004). Brown's point is the following: if there were such things as self-evident propositions, then why have internalists never appealed to the notion of self-evidence to resist the problem of forgotten evidence? After all, if self-evident propositions were a genuine possibility, then the internalist could allow for S's belief that p being justified at time t2 by saying that S's evidence for p at t2 is p itself. And yet, as Brown notices, "instead of attempting to argue that [S]'s justified belief that [p] is evidence for itself, [internalists] have attempted to identify other things which are part of [S]'s evidence and can support the target belief" (Brown, 2018, p. 53). For instance, they have cited the fact that S remembers coming to believe that p in a reliable way in the light of good evidence (See Conee & Feldman, 2004;McCain, 2014). According to Brown, the fact that no internalist has ever appealed to self-evidence speaks against the plausibility of self-evident propositions altogether.
Appealing to the distinction between evidence-for M and evidence-for J allows us to see that there are independent reasons why evidentialist-internalists have never appealed to self-evident propositions in order to resist the problem of forgotten evidence. Therefore, the fact that the notion of self-evidence is not considered in this debate cannot be used as a reason for thinking that there are no such things as self-evident propositions in any sense, and therefore it cannot be used as a way to show the implausibility of SKSS-J and infallibilism.
First, note that the problem of forgotten evidence was originally raised as a challenge for an internalist-evidentialist account of doxastic justification. To resist the problem of forgotten evidence, the evidentialist has to explain how S's belief that p at time t2 "can qualify as being doxastically justified under their theory" (Goldman, 1999(Goldman, , 2011. To do so, she has to identify the evidence S's belief that p at time t2 is based on. With this in mind, we can ask: if there were such things as self-evident propositions, can internalists resist this problem by saying that p is evidence for p? 15 I believe they cannot. Given the distinction between evidence-for M and evidence-for J , we can distinguish between the following two readings of self-evidence: Self-evidence M : p is motivating self-evidence iff p is the evidence on the basis of which one believes that p.
Self-evidence J : p is justifying self-evidence iff p stands in an evidential-support relation towards p so as to make p justified.
Perhaps the internalist could say that p is self-evident J insofar as p plausibly evidentially supports p (e.g., the probability of p given p trivially meets any given threshold required for evidential-support). Crucially, this would not help the internalist. For given the problem of forgotten evidence targets the internalist account of doxastic justification, the internalist will have to explain how S's belief that p at t2 is justified. To successfully overcome the problem, the internalist cannot merely point out to an evidential relation between an instance of evidence (p) S has and the target believed proposition (p). 16 Instead, she has to tell us which evidence S's belief that p at time t2 is based on.
Would appealing to self-evidence M work? Maybe the evidentialist-internalist could say that p is the evidence on the basis of which S believes that p, and that is what makes S's belief justified. This, however, would not be a satisfying reply. For what underpins these traditional internalist-evidentialist accounts of doxastic justification is exactly the assumption that there should be a logical space between a proposition p and the evidence on the basis of which one believes that p. As Conee and Feldmanthe main defenders of internalism-evidentialismsay when discussing the importance of a well-founded belief in doxastic justification: The term "well-founded" is […] used to characterize an attitude that is both well-supported and properly arrived at [Italic is mine]. Well-foundedness is an evidentialist notion because its application depends on two matters of evidencethe evidence one has, and the evidence one uses in forming the attitude. (Conee & Feldman, 2004, p. 93) 15 Moreover, note that Conee and Feldman explicitly endorse a view on which evidence is constituted by mental states, for example, experiences. Therefore, they would not even allow the proposition p to count as evidence. 16 Moreover, note that traditional evidentialists like Conee and Feldman (2004) and McCain (2014) embrace an explanationist account of evidential support, on which e is evidence for p iff p is the best explanation for e. However, it is not clear that p would thus in fact qualify as the best explanation for p itself, thereby failing to adequately stand in an evidential support relation towards p. More on this in the next section.
Appealing to self-evidence M would thus amount to saying that p is the evidence S uses at time t2 in forming the belief that p. But then, the problem with appealing to self-evidence M in forgotten evidence is not that this move leads to an infelicitous claim but rather that it leads to a false claim about what evidence one used as basis in coming to believe that p at t2.
Strategy 2: My diagnosis of why appealing to self-evident propositions would not be a successful move applies, mutatis mutandis, to the second strategy Brown considers, namely, the appeal to Gricean pragmatics. Let us see why.
According to strategy 2, in order to avoid the infelicity challenge, the infallibilist could insist that it is literally true to cite p as evidence for p while acknowledging that it is infelicitous to do so. Given that it would violate one or more Gricean conversational norms, it would be pragmatically inappropriate (Williamson, 2000, pp. 187-188). Crucially, according to Brown, this strategy will also fail. By inviting the reader to consider cases like Risk, Brown argues that there is an important disanalogy between settings involving a request for evidence for p and traditional cases in which something uttered is literally true but conversationally inappropriate. For in the former cases, "the problem of citing e as evidence for itself is not that this claim is obviously true! We struggle to see how it could be true" (2018,54). Remember the friend's response to the evidencerequest in Risk: "'alcohol is a cancer risk' is evidence for the claim that alcohol is cancer risk, but I don't mean to imply that I haven't got any other evidence for the claim that alcohol is a cancer risk" (2018,57). According to Brown, if we were in "Gricean pragmatics territory", then this sentence would pass the socalled cancellation test. That is, we would be able to cancel the second conjunct and be left with a claim that although inappropriate is nevertheless true. And yet, Brown argues, the first conjunct seems just false rather than merely inappropriate. Hence, the pragmatic response fails.
Once the distinction between evidence-for M and evidence-for J is available, we can make sense of why "we struggle to see how it could be true" (54) to cite p as evidence for p. The examples Brown considers, namely, those settings where "a request for evidence for p is a request for evidence for p constituted by propositions other than p, or 'non-p evidence'" (55), for example, examination settings, academic enquiry, and courtroom proceedings, suggest that what is infelicitous about the Infelicity data is the fact that in those settings one cites p as evidencefor M p. When we intuitively judge the truth conditions of such evidence-for statements in settings of enquiry, we are judging the truth conditions of statements about the evidence that convinced S to believe that p. This explains why the cancellation test does not work in cases involving evidence-for statements: the first conjunct is in fact false and not merely inappropriate, given it is false that S used the fact that alcohol is cancer risk as evidence for coming to believe that alcohol is a cancer risk. After all, Brown describes Detective Morse as a case in which "the detective Morse comes to know by eyewitness testimony that […] Burglar Bill was in the vicinity of the Central Jewellery Store just before the theft" (40) (italics are mine).
Where does this leave us? In the previous section, I have argued that the infelicity challenge rests on an ambiguity of the notion of the evidence-for relation. On the one hand, it is plausible to take infallibilism as a claim about evidential support. This means that all the infallibilist has to do to avoid scepticism is to endorse SKSS-J, and yet it is SKSS-M that underpins the infelicity data. In this section, I have shown that by appealing to the evidence-for M /evidence-for J distinction, and the corresponding SKSS-M/SKSS-J, we can see that strategy 1 and strategy 2 fail for reasons that are independent of infallibilism. Therefore, the failure of these strategies cannot be used, as Brown does, as an indication of the implausibility of infallibilism and, more precisely, of SKSS-J.

Objections
Before concluding this paper, I want to briefly address and resist two objections to my argument.
Objection 1: "Given the defender of Infallibilism wants to give an account of knowledge, she is also interested in doxastic justification. Therefore, infallibilism doesn't merely involve the justifying understanding of the evidence-for relation, but its motivating understanding as well. If she wants to avoid scepticism she will also have to endorse SKSS-M." 17 Response: First, note that probability 1 infallibilism merely gives us a necessary condition for knowledge. By doing so, it leaves it open how we should understand other necessary conditions on knowledge, for example, doxastic justification. Probability 1 infallibilism is compatible with the orthodox account of doxastic justification, one on which one's belief that p is doxastically justified only if one believes that p on the basis of one's propositional justification for p, but it is also compatible with a knowledge-first account of doxastic justification, one on which one's belief that p is doxastically justified iff one knows that p (e.g., Williamson, forthcoming;Sutton, 2005). In fact, infallibilism is compatible with the orthodox view on which knowledge entails doxastic justification, as well as with less orthodox views on which knowledge does not require doxastic justification (e.g., Foley, 1987). Even if we grant that the infallibilist is also concerned with doxastic justification, signing up to probability 1 infallibilism does not force the infallibilist to embrace SKSS-M or any specific account of doxastic 17 Thanks to Jessica Brown for raising this kind of objection to me in conversation. (Wording is mine) justification. However, the discussion in the previous section allows us to identify another version of infallibilism that we should expect to be problematic: Motivating infallibilism (MI): If S knows that p, then there is a proposition e, such that S believes that p on the basis of e and P(pje) =1 18 If the defender of MI wants to avoid scepticism, then she has to embrace SKSS-M: If S knows that p, then p is the evidence on the basis of which S knows that p. But what the above discussion teaches us is that paradigmatic cases of knowledge based on evidence are cases in which the evidence e that works as rational basis for believing (and knowing) that p is such that p ≠ e. Interestingly, MI faces a problem that is even more serious than the one allegedly posed by Brown's infelicity challenge. If we assume anti-scepticism, motivating infallibilism does not merely entail an infelicitous claim, but it entails a false claim about what constitutes one's rational basis in coming to believe and know that p. However, given what I have said above, it is clear that probability 1 infallibilism does not entail motivating infallibilism. As things stand, these challenges do not threat probability 1 infallibilism.
Objection 2: "Even if we grant that infallibilism entails SKSS-J (and not SKSS-M), the claim that p is evidence for itself in a justifying sense is equally infelicitous. Imagine that I explicitly ask the judge what justifying evidence there is for Smith being the murderer. It would be very weird if the judge responded that 'Smith being the murderer' is evidence for 'Smith being the murderer'." 19 Response: Once we interpret probability 1 infallibilism as a claim concerning evidential support and propositional justification, and once we appreciate the fact that propositional justification is merely about the logical relation between one's evidence and a target proposition p, I suspect the infallibilist will have few qualms in insisting that there is nothing infelicitous in saying that a proposition p is propositionally justified by itself. For this would amount to saying that p is trivially logically entailed by itself or that the probability of p given p is 1. However, I imagine some readers would not be entirely satisfied with this response. Fortunately, the infallibilist who is not willing to give up on his view but who also wants to side-step the justifying version of the infelicity challenge can do so. For, on closer inspection, we can see that not only is infallibilism not committed to SKSS-M, but, despite first appearances, it is not committed to SKSS-J either. To see why this is the case, let us recall what variety of infallibilism Brown is considering: Probability 1 infallibilism: If S knows that p, then the probability of p given S's total evidence E is 1.
As mentioned in the previous section, it is plausible to take probability 1 infallibilism to be about what kind of propositional justification (and evidential support) knowledge requires. In particular, it is plausible to take infallibilism to be a claim that rests on a probabilistic account of evidential support, for example, one on which one's total evidence E supports p iff E makes p sufficiently likely. However, this is not something that a defender of probability 1 infallibilism has to do. Note that, as Brown herself explicitly points out, one of the reasons we should take the infelicity challenge for P1I seriously is that P1I is entailed by a popular contemporary view about evidence, that is SKE: the thesis that if one knows that p, then p is part of one's evidence. Crucially, SKE is a claim about the nature of evidence, and it is by itself completely silent on what it means for evidence to support a proposition. That is, it is silent on how to spell out the evidential-support relation. And yet, the version of the infelicity challenge that (allegedly) still arises for infallibilism rests on the idea that infallibilism is committed to SKSS-J: if one knows that p, then p is part of S's (justifying) evidence for p. In particular, for the argument to work, it is crucial that we understand "p is part of S's evidence for p" correctly. For the argument against infallibilism will be effective if and only if infallibilism is committed to the idea that if one knows that p, then p provides adequate evidential support to p so as to make p propositionally justified. But once we understand that SKE is what generates P1I, we can see that the infallibilist can hold both SKE and P1I while endorsing an alternative non-probabilistic account of evidential support, one which does not give rise to the infelicity challenge. For instance, one could adopt SKE (and thus P1I), while endorsing an explanationist account of evidential support, one on which e is (justifying) evidence for a proposition p iff p is the best explanation for e (e.g., McCain, 2014). Insofar as what counts as best explanation will depend on various factors, such as, simplicity, explanatory power, and fruitfulness, it is plausible to say that p will not constitute the best explanation for p. But then it looks like P1I does not entail SKSS-J after all. Why does this matter? It matters because this puts the infallibilist in a position to explain why citing the proposition "Smith is the murderer" as evidence for the proposition "Smith is the murder" is not merely infelicitous but false even on an infallibilist account. For although the proposition "Smith is the murderer" will be part of the judge's evidence as soon as she comes to know that Smith is the murder (as entailed by SKE), the evidence "Smith is the murderer" is not evidence for the proposition "Smith is the murderer" because this fails to be the best explanation of why the proposition is true in the first place.
Another option available to the defender of SKE and P1I is to avoid SKSS-J by endorsing a version of pragmatic or moral encroachment. That is, one could say that whether evidence e offers adequate evidential support for a proposition p depends on the practical or moral stakes involved in a specific context (cf. e.g., Moss, 2018, Basu, 2019. 20 What this shows, once again, is that SKE and P1I are claims about the nature of evidence, whereas SKSS-J underpinning the justifying version of the infelicity challenge is a claim about what it takes for evidence to support a proposition. Let us recapitulate. I have pointed out that 1) P1I is a consequence of SKE, 2) and that SKE is a claim about the nature of evidence, thereby remaining neutral on what it means for an instance of evidence e to support a proposition p. Crucially, the justifying version of the infelicity challenge is supposed to arise for P1I only if P1I entails SKSS-J: if S knows that p, then p is justifying evidence for p. But once we realize that P1I merely states a necessary condition for knowledge, the infallibilist can endorse a non-probabilistic account of evidential support, for example, an explanationist account, without having to commit itself to the further claim that p is (justifying) evidence for itself. In other words: P1I does not by itself entail SKSS-J. Therefore, the justifying version of the infelicity challenge does not arise.

Conclusion
In this paper, I have argued that the infelicity challenge for infallibilism rests on an ambiguity of what it means for evidence e to be evidence for a proposition p. On the one hand, the infelicity challenge has been presented as the challenge of explaining the infelicity data, namely, that it is infelicitous to cite p as evidence for itself. However, the plausibility of the infelicity data rests on a motivating reading of the evidence-for relation. On the other hand, the thesis allegedly entailed by the infallibilist, namely, that if one knows that p then p is evidence for itself, is one that might be thought of as involving a justifying notion of evidence-for. Putting things in terms of justification, although it is plausible to think of infallibilism as saying something about propositional justification, it surely does not entail any specific account of doxastic justification. Moreover, I have shown that, although it is plausible to read infallibilism as entailing something about propositional justification, as it is defined, it is actually silent on how to understand the evidential-support relation (and thus propositional justification). Therefore, even if one insisted that there is another justifying version of the infelicity challenge the infallibilist has to resist, the infallibilist can do so because the justifying version of the infelicity challenge arises only if the infallibilist is committed to a probabilistic account of evidential support, one which entails SKSS-J: if S knows that p, then p is part of S's justifying evidence for p. However, I have shown that infallibilism, as defined, is perfectly compatible with other accounts of evidential-support, such as the explanationist account, which instead does not entail SKSS-J.
The conclusion we can draw is that, once it is clear that the infelicity data is about evidence-for M , the infallibilist can: accept the data, say that it is not literally true to cite p as evidence-for M p, while 1) either insisting that it is literally true to say that p is evidence-for J p (if they want to retain a probabilistic account of evidential support on pain of facing a justifying version of the infelicity challenge), or 2) insisting that it is also not literally true to cite p as evidence-for J p (e.g., if they endorse a non-probabilistic account of evidential-support, thereby resisting the justifying version of the infelicity challenge).