Facts about incoherence as non-evidential epistemic reasons

Abstract

This paper presents a counterexample to the principle that all epistemic reasons for doxastic attitudes towards p are provided by evidence concerning p. I begin by motivating and clarifying the principle and the associated picture of epistemic reasons, including the notion of evidence concerning a proposition, which comprises both first- and second-order evidence. I then introduce the counterexample from incoherent doxastic attitudes by presenting three example cases. In each case, the fact that the subject’s doxastic attitudes are incoherent is an epistemic reason to suspend, which is not provided by evidence. I argue that this incoherence fact is a reason for the subject to take a step back and reassess her evidence for her conflicting attitudes, and thus a reason to suspend all of them. Suspending judgment enables the subject to revise attitudes where appropriate and thus (typically) to arrive at a set of coherent and well-supported attitudes. I then address a dilemma for my proposal and, in conclusion, briefly suggest a picture of epistemic reasons on which they are to be understood against the background of the subject’s virtuous intellectual conduct.

1 Introduction

A recent trend in epistemology has been to explain epistemic phenomena or to analyze epistemic concepts by appealing to epistemic reasons, while taking inspiration from debates about reasons for action, for instance with respect to the metaphysical nature or normative and explanatory roles of reasons. This trend is most clearly exemplified by reason-first approaches to epistemology (Schmidt, 2021; Schroeder, 2021).

If epistemic reasons are to take center stage in epistemology, however, we need to get clear on how they relate to evidence. Evidence features in many long-standing epistemological controversies, e.g., over justification, the nature of epistemic obligations, or the epistemic implications of peer disagreement. At the same time, epistemic reasons and evidence play many of the same roles, and moreover, some epistemologists have claimed that epistemic reasons and evidence come down to the same thing (Kelly, 2014; McCain, 2014; Schroeder, 2008). So, arguably, talk of evidence may be replaced by talk of reasons without loss. This could be problematic for reasons advocates, for another way of putting this point is: Epistemologists will not miss out on anything if they keep on talking about evidence; current attempts to move our focus to epistemic reasons are unlikely to provide insight beyond that provided by traditional evidentialism.

Against this backdrop, the aim of this paper is to present a counterexample to the following claim:

• (R-E) All epistemic reasons for doxastic attitudes towards p are provided by evidence concerning p.

For instance, my reason to believe that it will rain tomorrow is provided by the weather forecast that predicts rain for tomorrow, i.e., by evidence that it will rain tomorrow. (R-E) expresses an evidentialism about epistemic reasons. It leaves it open whether epistemic reasons are to be identical with evidence, grounded in evidence, or something else still—I take it that the term “provide” is compatible with a broad range of possible relations. In this way, my counterexample to (R-E) aims to cause trouble for evidentialism about epistemic reasons quite generally. (Note that (R-E) does not commit its adherents to evidentialism as the view is sometimes understood: the claim that there are no practical, but only evidential reasons to believe (Shah, 2006). Even if all epistemic reasons are provided by evidence, there may be further, non-epistemic reasons that bear on doxastic attitudes.)

How are “epistemic reasons” and “evidence” to be understood? I focus here on epistemic normative reasons, i.e., considerations that favor adopting a certain doxastic attitude. Considerations may be thought of as propositions, and true propositions as facts. Epistemic normative reasons are considerations that favor a subject’s doxastic attitudes in an epistemic rather than a practical manner.¹ I speak of doxastic attitudes rather than belief to include epistemic reasons that count in favor of disbelieving p and for epistemic reasons to suspend judgment on whether p. Reasons to suspend judgment are reasons to adopt an attitude of being unsettled with regard to the question of whether p, that is, a question-directed attitude (Friedman, 2013, 2017).

“Evidence” is a multi-faceted term (see Kelly, 2014). Among other things, Kelly distinguishes the theoretical roles of evidence, of making doxastic attitudes reasonable, and of indicating truth. I will assume here that evidence is that which justifies doxastic attitudes by indicating the truth or falsity of their contents.²

Epistemic reasons as well as evidence come into their own when they are had by a subject—only then do they plausibly make a difference to the epistemic status of her doxastic attitudes (Schmidt, 2019). When we inquire into whether a certain attitude is epistemically justified for a subject, we naturally turn to the reasons or evidence possessed by the subject. Accordingly, I will leave to one side unpossessed epistemic reasons or evidence. Reasons (evidence) can bear on the justification of a doxastic attitude either as pro tanto reasons (evidence) or as all-things-considered reasons (evidence). In the first case, they are factors that merely contribute to the justification of a doxastic attitude, without fixing its epistemic status; in the second case, they include all the reasons (evidence) that a subject possesses concerning a certain matter, which together settle whether and how strongly the attitude is justified.

The counterexample I present, then, shows the falsity of the claim that all normative epistemic reasons—considerations that favor a doxastic attitude in a characteristic epistemic way—are provided by evidence, which bears on the justification of a doxastic attitude by indicating the truth or falsity of its content.

Here is how I will proceed: I begin by showing why (R-E) may initially seem compelling. One might argue that the principle is true by definition, in virtue of the meanings of “epistemic reason” and “evidence” (Section 2). I then look at two ultimately unsuccessful counterexamples to (R-E), one appealing to facts about evidence and the other to the fact believed itself (Section 3). This will help us to gain a clearer understanding of (R-E) and the associated picture of epistemic reasons. In Section 4, I provide the counterexample that is the focus of this paper by presenting three cases of incoherent or inconsistent doxastic attitudes, and I argue that these cases involve non-evidential epistemic reasons. In Section 5, I address a dilemma for my proposal. I conclude by gesturing at an alternative picture of epistemic reasons, which can make sense of the cases I present as cases of epistemic reasons. In this picture, all epistemic reasons can be understood against the background of an agent’s virtuous intellectual conduct.

2 The case for (R-E)

The strongest motivation for (R-E) is that it may seem to be true by definition. Evidence is that which bears on the justification of a doxastic attitude, in a characteristically epistemic or truth-related way (Goldman, 2011, 255; Kim, 1988; Kelly, 2014). But exactly the same can be said about epistemic reasons (Gibbons, 2010, 344). Given that bearing on the justification of a doxastic attitude is what defines both evidence and epistemic reasons, it appears that it has to be evidence which provides epistemic reasons.

Moreover, one might argue that (R-E) best explains what is epistemic about epistemic reasons, i.e., the characteristically truth-related way in which epistemic reasons favor doxastic attitudes. Since there are arguably practical reasons for doxastic attitudes in addition to epistemic reasons, it is a genuine question what makes a doxastic attitude epistemically rational and, relatedly, what distinguishes epistemic reasons from practical reasons. From Cohen’s (1984) discussion of the truth-connection, we can take the idea that epistemic reasons favor a belief in the virtue of bearing on its truth. Plausibly, to have an epistemic reason to believe that p is to be in possession of a consideration that favors adopting the belief that p by indicating, guaranteeing, or otherwise bearing positively on the truth of p. A practical reason for a doxastic attitude, by contrast, counts in favor of the attitude by showing that it is good or useful to have, while being silent on the truth or falsity of its content.

Here is how (R-E) explains the power of epistemic reasons to favor epistemically. Evidence, as something that indicates that a proposition is true (or false), seems to have just the kind of connection to truth that is essential to the favoring done by epistemic reasons. Epistemic reasons count in favor of belief and the other doxastic attitudes by bearing on the truth of their contents, while evidence justifies doxastic attitudes by indicating the truth or falsity of their contents, i.e., (also) by bearing on their truth. So, the best explanation of how considerations epistemically favor a doxastic attitude towards p is that they are provided by evidence concerning p.

The picture of epistemic reasons associated with (R-E) that has emerged is this: a consideration epistemically favors adopting a doxastic attitude with a content p when it bears on the truth of p, i.e., when it is provided by evidence concerning p. Importantly, this is not the unrestricted claim that any epistemic reason is provided by evidence for something or other, which I think is true, but trivially and uninterestingly so. An epistemic reason is a (true) proposition. It is unsurprising that we can always find something of which a (true) proposition is evidence. For instance, the fact that I have no evidence concerning the question whether it will rain tomorrow is a reason to suspend. And it is certainly itself evidence of something, for instance, evidence that I have not checked the weather forecast today. This is beside the point, however. The contentious question in the example is: is my reason to suspend on whether it will rain tomorrow provided by evidence relevant to the truth or falsity of the claim that it will rain tomorrow?

This gets us some of the way towards clarifying (R-E), though we still need to pin down what exactly is involved in evidence “concerning” a proposition or “being relevant” to its truth. To do so, let us now turn to two ultimately unsuccessful counterexamples to the principle.

3 Some inconclusive counterexamples

Some counterexamples to (R-E) have been proposed in the literature, which—it is claimed—involve epistemic reasons that are not provided by evidence. Let us start with the counterexample from facts about evidence.³ Here are some cases: the fact that the subject’s evidence concerning p is exactly balanced—or that she has zero evidence concerning p—is a reason for her to suspend judgment on whether p, but it is not itself a piece of evidence indicating that p is (or is not) the case. Next, say a subject now has some evidence indicating that p, which would normally suffice for the justified belief that p. However, further evidence concerning p is forthcoming. This fact is not itself a piece of evidence that p (or that not-p), but in light of it, it seems that suspension is the right attitude towards p. Plausibly, this fact is a reason to suspend that is not provided by evidence concerning p (Schroeder, 2012a, 284).

Other cases involve inference to the best explanation or induction. In inference to the best explanation, the subject’s evidence consists in the data for which we are trying to find an explanation. That dinosaur fossils and modern-day birds share features such as feathers is best explained by the hypothesis that birds are feathered dinosaurs. The subject’s evidence consists in facts about features of dinosaur fossils and birds. The fact that the shared features of dinosaur fossils and of modern-day birds are best explained by the hypothesis that birds are feathered dinosaurs is not part of the evidence, but a fact about what best explains the evidence. Yet, it is an epistemic reason to believe that birds are feathered dinosaurs. Similarly for induction: so far, all of my morning experiences were of the sun rising. It is a fact, then, that my evidence inductively supports the belief that the sun will rise tomorrow. This fact itself is plausibly a reason for me to believe that the sun will rise tomorrow, but not a piece of evidence in support of my belief; after all, it is a fact about all my evidence in support of the belief. Quite generally, the fact that I have evidence that p is a reason to believe that p. But since it is a fact concerned with my evidence, not with the issue of p itself, it seems that it is not itself evidence that p.

The counterexample presupposes that evidence concerning p is limited to evidence that speaks directly to the truth or falsity of p, that is, first-order evidence. Proponents of (R-E) can easily reject this assumption. It is not only pieces of first-order evidence that count as considerations that bear on whether p. The fact that the subject has exactly balanced first-order evidence concerning p, that conclusive first-order evidence concerning p is forthcoming, that the subject has evidence concerning p, and so forth, is higher-order evidence concerning p—evidence concerning S’s evidence that p, and thereby also evidence concerning p. In general, it is plausible that higher-order evidence of evidence that p is a reason to believe that p; higher-order evidence of evidence that not-p is a reason to disbelieve p; and higher-order evidence that the subject’s evidence concerning p does not settle the issue is a reason to suspend on p.⁴

A different case of higher-order evidence has been proposed by Christensen (2010). A pilot in flight calculates correctly that she has sufficient fuel to get to her destination. She then learns that there is a 0.75 chance that she is suffering from hypoxia, a condition that interferes with people’s ability to do arithmetic. That there is a 0.75 chance that she is suffering from hypoxia is evidence that she is unable to calculate her fuel levels correctly; it is, plausibly, a reason for her to suspend judgment on whether she has sufficient fuel. However, this is evidence concerning her cognitive functioning rather than evidence of evidence, as in the previous cases. Should we say that this is still higher-order evidence concerning p?

Proponents of (R-E) can argue for a positive answer to this question: In trying to settle an issue, evidence concerning one’s own cognitive functioning plays a central role. When the subject realizes that her cognitive faculties involved in figuring out whether p are impaired, this serves as a defeater—specifically, a defeater undercutting her evidence concerning p. In an analogous case, in which the subject realizes that one of her informants with respect to p has impaired cognitive abilities, this is not evidence that directly bears on the issue of whether p either. Still, it serves as a defeater of the evidence provided by the informant, and so is evidence that is relevant to whether p for the subject. For the hypoxia case, we can also put it like this: That the pilot’s arithmetic capabilities are probably impaired is evidence that she failed to settle the issue whether she has enough fuel to get to her destination. So it is evidence that the issue is not settled, and thus—unproblematically for (R-E)—provides a reason to suspend.

To sum up, cases of facts about evidence show that proponents of (R-E) should include, under evidence concerning p, not only evidence directly bearing on the truth or falsity of p, but also higher-order evidence—evidence about one’s evidence as well as evidence concerning one’s own cognitive processing of one’s evidence.⁵

Next, consider the counterexample from the fact believed itself, which can be used to motivate a probability-raising understanding of the evidence for proponents of (R-E). Austin (1961, 115) famously argued that, in cases where the facts simply present themselves to me—as when I have a pig plainly in view—it cannot be said that this provides evidence that they obtain. For instance, that there is a pig in front of me, or that I see the pig, is not evidence that there is a pig in front of me.⁶ In such a situation, it is simply established that there is a pig in front of me, so there is no more question of evidence of the pig’s presence at that point. But there is an epistemic reason in such cases. When the facts simply present themselves to me, as when I have the pig plainly in sight, I have every reason to believe that there’s a pig! What favors my belief that there’s a pig in front of me specifically, i.e., is a normative reason, is simply the fact that there’s a pig in front of me. Imagine my friend asking me, “Why do you believe there’s a pig in front of you?” (or even “How do you know there’s a pig in front of you?”). A natural response would be: “Just look, there it is!” pointing my friend to the relevant epistemic reason: the fact that there is a pig. If this is correct, then I have an epistemic reason to believe that p that is not provided by evidence concerning p.

Proponents of (R-E) might respond that in such cases, the subject believes without a reason (Littlejohn, 2018). I do not find this line of argument convincing, since it seems clear that in such cases there is something that counts in favor of believing, viz. the very fact believed (see Schmidt, 2021). I will instead focus on a line of response on behalf of (R-E) that claims that the fact that p is evidence that p after all, which takes on board a probability-raising conception of evidence.⁷ A piece of evidence e that p raises the probability that p vis-à-vis a given body of evidence b just in case the probability that p is higher given b + e than it is given b alone (Williamson, 2000, 186/187). In this view, the fact that there is a pig in front of me is evidence that there is a pig in front me. For this fact does raise the probability that there is a pig in front of me; if this fact is included in my evidence, the probability that there is a pig in front of me is 1. But excluding this fact from my evidence, the probability that there is a pig in front of me is lower than 1. So, my reason to believe that there is a pig can unproblematically be provided by evidence that there is a pig.

The take-home message from this section is, first, that proponents of (R-E) can resist counterexamples from facts about evidence if they understand evidence concerning p to include higher-order evidence regarding p and, second, that adopting a probability-raising conception of evidence allows them to avoid the alleged counterexample from the fact believed itself. With this, I turn to a kind of counterexample that causes trouble even for this amended version of (R-E), which involves incoherent doxastic attitudes.

4 The counterexample from incoherent attitudes

In this section, I will present three cases of incoherent or inconsistent attitudes: the case of a 6/49 lottery, a case of disagreement between Miss Marple and Hercule Poirot, and a case of history vs. philosophy.⁸ I will argue that in each of these cases, the fact of the incoherence of the subject’s attitudes (“incoherence fact,” for short) provides a reason for her to suspend on whether p that is not provided by evidence concerning p, and so that (R-E) is false.⁹

6/49 Lottery¹⁰

Lola participates in a lottery in which each player chooses six numbers from 1 to 49, and wins the jackpot if her numbers match the six numbers produced in the drawing. The probability that she will win the jackpot is 1 in 13,983,816. There is no guarantee that anyone will win the jackpot. The lottery has millions of regular participants, and as a matter of fact, the chance in any drawing that at least one player wins the jackpot is extremely high. To fix ideas, say that over the last five decades, there has been only 1 in 1000 drawings in which no ticket won the jackpot. Lola is aware both of the extremely high chances of losing of every single ticket and, by way of induction, of the extremely low chances that everyone’s tickets will lose (not win the jackpot). It is then extremely probable, from Lola’s point of view, for ticket 1, that it will lose; for ticket 2, that it will lose; …; for ticket n, that it will lose (call the respective propositions “p₁,” “p₂,” …, “p_n”). But at the same time, it is extremely probable for her that it is not the case that ticket 1 will lose and that ticket 2 will lose, …, and that ticket n will lose (call this proposition “p_□”).

In 6/49 Lottery, Lola has sufficient evidence for each of p₁, p₂, …, p_n, and p_□, and so, according to (R-E), sufficient reasons to believe each proposition. Given that sufficient reasons render the belief justified, she has justified beliefs p₁, p₂, …, p_n, and p_□. However, these beliefs are jointly inconsistent: It cannot at the same time be true, of each ticket, that it will lose and not be the case that each ticket will lose.

Lola has jointly inconsistent, yet individually justified beliefs. Assume that she realizes that her beliefs are inconsistent. Now, it is epistemically irrational for her to have inconsistent beliefs, especially given that she is aware of their inconsistency. So she cannot be justified in holding all of these beliefs; that is, it cannot be that each of them is justified (cf. Nelkin, 2000, 375).

In light of this, I suggest that the fact that Lola’s beliefs are inconsistent is a pro tanto epistemic reason for her to suspend on each of p₁, p₂, …, p_n, and p_□. As soon as she becomes aware of the inconsistency, she possesses this reason to suspend. It is a reason to take a step back from the whole mess that is her beliefs and to rescind her endorsement of every single proposition that gives rise to the inconsistency, i.e., to suspend on all of them. When a subject stumbles upon an inconsistency in her beliefs, this is a reason she has to reassess her whole epistemic situation with respect to the issue; to do so, she has to reopen the questions that she previously settled by believing the relevant propositions, but that is just to say that she has to suspend on these propositions. So she has a reason to suspend on all the beliefs which give rise to the inconsistency.¹¹

By contrast to the hypoxia case sketched above, Lola’s reason to suspend—the fact that her beliefs are inconsistent—is not an undercutting defeater vis-à-vis her reasons to believe p₁, p₂, …, p_n, and p_□. It does not bear on her extremely strong evidence concerning p₁, p₂, …, p_n, and p_□—the respective probabilities of someone particular/of anyone winning the jackpot are untouched by the inconsistency of Lola’s beliefs. Rather, it is a reason that favors an attitude that is incompatible with belief, viz. suspension. (Contrast this with the hypoxia case, where the fact that the pilot’s cognitive processing may be malfunctioning undermines her evidence whose production relied on this cognitive processing.)

Furthermore, the fact that Lola’s beliefs are inconsistent does not make a difference to the all-things-considered probability of their truth, given her body of evidence. The probabilities that the propositions are true, determined by doing the math and by doing induction, remain unchanged.¹² It is not itself a piece of evidence concerning p₁, p₂, …, p_n, or p_□, not even a piece of evidence that Lola’s evidence does not settle for her whether p₁, p₂, …, p_n, or p_□. For, on the one hand, the fact that these beliefs are inconsistent all by itself does not tell us anything about what her evidential situation is with regard to any one of them. On the other, each of her beliefs was formed by perfectly well-functioning processes and in response to very strong evidence, so overall her evidence does plausibly settle whether p, for each of them. We can even imagine that Lola is aware of the lottery paradox and knows that it is not any cognitive malfunctioning on her part that is at fault for the inconsistency. Nor do other facts about the subject’s evidence concerning these propositions ground the inconsistency—inconsistency is a matter of the relations between the propositions believed, not of the evidence for them. So the incoherence fact is not higher-order evidence concerning the relevant propositions. So, contra (R-E), this is a case of an epistemic reason to suspend on whether p that is not provided by evidence concerning p.

Lottery cases involve the extra twist that reassessing one’s situation will not make the inconsistency go away. Lola’s evidence for every single proposition is exceedingly strong, so there is no way for her to revise her beliefs in the light of her evidence-based reasons so as to resolve the inconsistency. So all things considered, she has—unchangingly—a reason to suspend on every single proposition involved. Her reason to suspend outweighs her evidence-based reasons to believe any of the propositions that give rise to the inconsistency. This also explains why—as is widely accepted—subjects in lottery cases do not justifiably believe that their tickets will lose.

There are many, more pedestrian cases of incoherent doxastic attitudes that involve no paradoxes, but incoherence-based reasons to suspend. One such case, modified from Worsnip (2018), is the following case of epistemic akrasia:

Marple and Poirot

Miss Marple and Hercule Poirot team up investigating a murder. Master detective Miss Marple is first on the scene and takes in all the evidence, forming the (for once, mistaken) belief that the evidence indicates that the vicar did it (v), and she tells Poirot so. That Miss Marple provides this testimony is a sufficient reason for Poirot to believe that the evidence indicates that v, and he forms the belief for that reason. Next, Poirot himself takes in the evidence at the crime scene, which as a matter of fact indicates that the vicar did not do it; he therefore has sufficient reason to disbelieve v, and disbelieves v for that reason. Poirot now has incoherent doxastic attitudes, belief that the evidence indicates that v and disbelief that v. They are incoherent because by virtue of his belief about the evidence, he accepts that there is sufficient evidence and thus reason to believe that v, and thus that belief that v is the correct doxastic response; but nonetheless, he disbelieves v.

In this scenario, both attitudes are supported by sufficient evidential reasons and justified on that basis. Upon finding himself with incoherent doxastic attitudes that are once due to Miss Marple’s expertise (the belief that the evidence indicates that v) and once to his own (disbelief that v), Poirot may with equal justification worry that Miss Marple has for once been mistaken or that he himself has misinterpreted the evidence. The fact that Poirot’s doxastic attitudes are incoherent is a pro tanto epistemic reason for him to take a step back from the whole mess and to reassess his epistemic situation—i.e., a reason to suspend on both propositions. Imagine that a moment after he has come to disbelieve that v in response to the evidence on the scene, he becomes aware that he has two incoherent attitudes, so that he possesses the incoherence fact as a reason to suspend on both. In light of his reason, he is plausibly justified to suspend both on whether the evidence indicates that v and on whether v, with an eye to figuring out how to resolve the tension between his attitudes. So his reason to suspend on the incoherent attitudes outweighs the evidential reasons backing these attitudes. It is sufficient for him to suspend—at least as long as he has no full view of his evidence-based reasons and their quality. (I elaborate on the temporal dimension of incoherence-based reasons below.)

As in 6/49 Lottery, the incoherence fact here is not itself a pertinent piece of first-order evidence. That his belief and disbelief are incoherent is no first-order evidence pertinent to the propositions in question, it neither indicates that the vicar did it (against his disbelief that v), nor does the incoherence fact directly indicate that the evidence does not support that the vicar did it (against his belief that the evidence indicates that v). Concerning the probability-raising conception of evidence: Adding the fact that the two attitudes are incoherent to Poirot’s evidence does not bear on the probabilities of the relevant propositions. As before, if the tension between disbelieving v and believing that the evidence indicates that v affects the probabilities of the respective propositions, it does so prior to adding the incoherence fact to his evidence.

Moreover, the incoherence fact is not higher-order evidence. That his belief and his disbelief are incoherent is not a fact about his evidence, as in the examples in Section 3, but a fact about the relation between his doxastic attitudes. To see this, compare, e.g., the fact that the subject has zero evidence concerning p with the fact that Poirot’s belief and his disbelief are incoherent. The former fact is concerned with the subject’s evidence, whereas the latter fact is about the subject’s mental states.

Finally, the incoherence fact is not evidence that Poirot’s evidence does not settle whether the propositions contributing to the incoherence are true.¹³ It does raise the worry that not everything has been said regarding these matters, i.e., regarding whether the vicar did it and whether the evidence indicates that the vicar did it—for if everything had been said, one would expect a coherent worldview. Nonetheless, the incoherence fact does not determinately indicate for either proposition that Poirot’s evidence does not settle whether it is true. Nor does it indicate that the evidence settles neither issue. The mere fact of incoherence does not give the subject any indication as to which attitude exactly is at fault, and so also gives no indication as to whether it is one or the other issue that his evidence does not settle. Additionally, for Poirot, a highly competent and conscientious reasoner, the incoherence fact is not evidence of malfunctioning. His incoherent doxastic attitudes are due not to bad reasoning, but to misleading evidence, and so are not evidence that his reasoning processes failed to settle the relevant issues. So, unlike in the hypoxia case, this is not evidence undercutting Poirot’s relevant first-order evidence.

So, just like in 6/49 and against (R-E), in Marple and Poirot, the incoherence fact is an epistemic reason for Poirot to suspend that is not provided by his evidence concerning the relevant propositions. A difference to 6/49 is that when Poirot concludes his reassessment of his epistemic situation, he will certainly realize that disbelief is indeed the justified attitude with respect to v, given the evidence at the scene; he will conclude that Miss Marple must have been mistaken about what the evidence indicates. That is to say, he will be justified to believe that the evidence at the scene supports disbelieving that v; say he does so, thereby abandoning his previous belief that the evidence supports belief that v. Then, the incoherence will have disappeared and with it the original reason to suspend.

Unlike in lottery cases, in more pedestrian cases of incoherent doxastic attitudes, a reassessment of the subject’s epistemic reasons will often uncover that some of her attitudes, but not others, are supported by sufficient evidential reasons, so that a revision of the unsupported attitudes is called for. By responding to her reasons and revising, the subject will eliminate the incoherence and thus lose her reason to suspend. Moreover, irrespective of her reason to suspend, the subject will often all along have all-things-considered evidence-based reason only for one of the two attitudes, though she is in a position appropriately to respond to these reasons only once she has reassessed her epistemic situation (and her reason to reassess is at the same time her reason to suspend).

Let us now briefly turn to a third case:

History vs. Philosophy

When Basna studies history, she comes to believe, based on her professors’ arguments, that the historical facts are relative to the historian who interprets historical sources. Basna later switches her focus to philosophy and, based on her philosophy professors’ arguments, she forms the belief that no facts are relative to anyone. Both beliefs are supported by sufficient evidence, which was provided, respectively, by Basna’s history professors and by her philosophy professors. Her evidence comprises both the professors’ expert testimony and the arguments with which she engages. At some point, Basna realizes that the two beliefs are inconsistent.

As in the previous cases, the fact that the beliefs are inconsistent is a pro tanto epistemic reason for Basna to reassess how she acquired either one and to figure out which of them, if any, is supported by sufficient evidence. It is a reason for her to re-open the inquiry and to suspend on both beliefs. We can imagine that the arguments provided by Basna’s philosophy professors are superior to those provided by her history professors, but that, due to compartmentalization, she simply has not compared notes, so to speak, and so has not yet weighed the arguments against each other. Once Basna starts to consider the arguments on both sides and realizes that the arguments against relativism are stronger than those for the claim that the historical facts are relative, she will stop suspending on the latter claim and start to disbelieve it. The inconsistency will thus be resolved and the reason to suspend will have disappeared.

As before, Basna’s reason to suspend is not provided by evidence: The fact that it is inconsistent to believe both that historical facts are relative and that no facts are relative is not first-order evidence bearing on the truth or falsity of relativism. In case we want to say that the inconsistency of Basna’s beliefs lowers the probability of each of them to 0, Basna’s belief, that no facts are relative, already decreases the probability of her belief, that the historical facts are relative, to 0, and vice versa, before she comes to believe that these beliefs are inconsistent. So, adding the incoherence fact to her body of evidence does not affect their respective probabilities. Moreover, this fact is not relevant to higher-order evidence. It is not a fact about Basna’s evidence, but a fact about her mental states. And it is not evidence that her cognitive processing is malfunctioning as in the hypoxia case. It admittedly does indicate that she has not settled the question of whether relativism is true, but this leaves it open that the cause of the inconsistent beliefs is quite harmless—as in having a compartmentalized mind—and so it does not undermine her first-order evidence.

History vs. Philosophy is thus another instance of the counterexample from incoherent attitudes for (R-E). The fact that Basna’s beliefs are inconsistent is a reason for her to suspend on both beliefs, but it is not provided by evidence concerning the relevant propositions.¹⁴

My claim that the three cases just presented provide a counterexample to (R-E) rests on the following core features of incoherence-based reasons: (1) It is, on the one hand, crucial that these reasons are not provided by evidence concerning the relevant propositions, as introduced above. As pointed out above, it is no objection to the counterexample that any incoherence facts will be evidence of something. However, for the counterexample to work, they cannot be first-order evidence concerning the propositions in question, they cannot be facts about such evidence, and they cannot be undercutting defeaters with respect to the relevant propositions. And given that (R-E) is combined with the probability-raising conception of evidence, it is important to my case that incoherence-based reasons do not affect the probability for the involved doxastic attitudes. On the other hand, incoherence-based reasons need to be epistemic—which, prima facie, they are: They intuitively bear on the epistemic standing of doxastic attitudes, primarily on suspension of judgment. More needs to be said here, though, which I will do in the following section.

(2) I have argued that incoherence-based reasons are unlike the higher-order defeating reasons in hypoxia cases in that they do not work by undercutting the evidential reasons supporting the relevant doxastic attitudes. Instead, they outweigh these reasons. Incoherence-based reasons function not as reasons modifying the subject’s evidential reasons for the relevant doxastic attitudes, but they are reasons for an alternative, suspension. Since suspension and belief/disbelief regarding a certain proposition rationally exclude each other, these reasons not only weigh in favor of suspension, but can thereby also outweigh the subject’s evidential reasons for her given doxastic attitudes.

Here is an analogy to illustrate the point: Imagine that, on Friday at 8 p.m., Adem is about to enter a pizzeria. He has sufficient reason to eat at the pizzeria—he is hungry and the pizza is great. But then he remembers that he promised his friend that he would walk her dog on Friday at 8. This promise-based reason is, in the first place, a reason to turn around and go walk the dog now. Since Adem cannot both eat at the pizzeria and go walk the dog now (and he cannot rationally both intend to eat at the pizzeria and to walk the dog now), his reason to walk the dog thereby also counts against eating at the pizzeria now. Notably, his promise-based reason does not undercut his reasons to eat pizza now; instead, it is a reason to perform an alternative action. In the same way, each of my cases involves an incoherence-based reason to suspend that outweighs, rather than undercuts, the evidential reasons backing the relevant doxastic attitudes.

(3) I have claimed in each case that the incoherence-based reason bears on all involved doxastic attitudes equally.¹⁵ This is important to my account because the point of suspension in each case is to enable genuine inquiry into the relevant matters. Suspension is supported by the reason for all involved propositions, and (dis)belief needs to be abandoned with respect to each of them, so that the subject is able to reevaluate the evidential reasons backing each proposition, and in the end to settle on a coherent set of attitudes. The core idea is that genuine inquiry into a matter presupposes an open-minded attitude towards it, i.e., suspension; and that finding oneself with incoherent attitudes is a reason to re-open inquiry into all matters about which one had previously formed belief, and had thus taken as settled.

(4) Finally, incoherence-based reasons are agent-relative and have a specific temporal profile (Christensen, 2010; Staffel, 2021). Regarding the first point, consider Basna. That her beliefs about relativism are incoherent is a reason for her to suspend judgment on whether relativism is true, because she needs to re-open inquiry so as to resolve the incoherence in her belief system. But it is not, for instance, a reason for me to suspend on whether relativism is true. Say I have a well-grounded belief that relativism is false. If I then learn that Basna has inconsistent beliefs on the matter, this does not affect my evidence-based reasons for my belief. This fact may be a reason for me to believe that Basna should re-open the inquiry and that she should suspend, but since the reason to suspend is due to her incoherence and her need to re-open the question, it does not affect my reasons. It is an agent-relative reason.

Regarding the temporal profile of incoherence-based reasons, they remain only until the incoherence is resolved, since that is their point. They are reasons to form what Staffel (2019, 284) calls “transitional attitudes,” viz. attitudes that are formed at the beginning or during deliberation, before a question is settled. And the states of suspension in each case are exactly such transitional attitudes. In line with Staffel’s view, whether a subject has an incoherence-based reason to suspend depends on her reasoning stage, whether she has such a reason can vary even where her evidence is unchanged. In Basna’s case, throughout, the evidence she possesses all things considered supports the belief that relativism is false. Realizing that her beliefs are incoherent rightly gives her pause, and to get into a position to respond correctly to her evidential reasons, she first has to re-open the question and so to suspend, before she can form the justified belief that relativism is false.

Here is one concrete way in which the subject’s reasons may change in the course of her deliberation. Basna has, first, realized that it is inconsistent to believe that no facts are relative and to believe that the historical facts are relative, and this has—rightly—caused her to suspend. Second, she realizes that in one of her philosophy classes, she has encountered an iron-clad argument to the conclusion that no facts are relative. She can now perform a disjunctive syllogism: It cannot be true both that the historical facts are relative and that no facts are relative; no facts are relative; so it is not the case that the historical facts are relative. In this version of History vs. Philosophy, the inconsistency fact is first a reason to suspend and then, together with Basna’s evidential reason for her other belief, a reason to disbelieve that the historical facts are relative.¹⁶¹⁷

These four features of incoherence-based reasons highlight the ways in which—though epistemic—such reasons differ from epistemic reasons provided by first-order or higher-order evidence. My discussion so far should go a long way towards defending the counterexample from incoherent attitudes. Yet, it is still possible to raise objections. In the following section, I will focus on one central objection, which takes the form of a dilemma.

5 A dilemma for incoherence-based epistemic reasons

The dilemma for my proposal that there are incoherence-based epistemic reasons to suspend is the following. Either incoherence-based reasons are provided by evidence after all or, failing this, they are not epistemic reasons (contra core feature (1)). The worry is that—first horn—contrary to what I have claimed, incoherence-based reasons are evidential. However, if I succeed in showing that incoherence-based reasons are not provided by evidence, then this just goes to show that they are not epistemic. This is the second horn.

I begin with the first horn. A good way of arguing that incoherence-based reasons to suspend on p are provided by evidence concerning p is to present a concrete way in which they can be traced back to such evidence. Here are two especially challenging ways: First, the fact that the subject’s doxastic attitudes are incoherent is evidence that something has gone wrong in her reasoning. Just like in the hypoxia case, it is a reason that undercuts her evidential reasons for her attitudes. Second, the incoherence fact is a fact about evidence. It indicates that something was wrong with the subject’s evidence, and my cases should be treated in line with the other cases involving facts about evidence discussed above.

Regarding the first way, the three cases have been constructed so that they do not exhibit the suggested parallel with the hypoxia case. In that case, the fact that the pilot’s reasoning is likely faulty (as she knows) is a reason for her to discount her evidence that was processed by this reasoning. So the fact acts as an undercutting defeater. By contrast, Lola’s reasoning in 6/49 Lottery is not faulty, she is simply stuck in a lottery paradox, and she is aware of this. Poirot’s reasoning in Marple and Poirot is also perfectly fine, and we can imagine that he is aware that he has made no reasoning mistakes. Regarding History vs. Philosophy, there is a version of the case in which Basna simply has two conflicting pieces of testimonial evidence from two different sets of experts (her history professors and her philosophy professors). Again, imagine Basna is aware of this; then overall, the incoherence fact is no evidence that her reasoning is faulty. But still, in all three cases, the subjects have an incoherence-based reason to suspend (feature (2)).

But doesn't my claim that the subjects’ reasoning is flawless in my cases strengthen the second way of tracing incoherence-based reasons back to evidence, viz. that the incoherence fact is evidence that the subjects’ respective evidence is faulty? In 6/49 Lottery, this is not the case (as Lola knows)—each proposition is supported by very good probabilistic evidence, and still, Lola has a reason to suspend. That is why lottery cases are paradoxical. I concede that, as described, Marple and Poirot and History vs. Philosophy involve misleading evidence and that the fact that the subjects’ doxastic attitudes are incoherent is plausibly evidence that some of their evidence is misleading.

However, the incoherence facts do not acquire their status as reasons to suspend from being evidence of evidence.¹⁸ Lola, Poirot, and Basna all have a reason to suspend of the very same kind, which is due to how correct reasoning works: When a subject realizes she has incoherent attitudes, it is rational for her to re-open inquiry to the end of reaching a state in which all her attitudes cohere, and thus she has a reason to suspend. 6/49 Lottery illustrates that this holds independently of whether the subject’s incoherent attitudes also indicate that something is wrong with her evidence. As elaborated above (core feature (2)), reasons of this kind do not function by undercutting the subject’s evidential reasons (as does evidence that something is wrong with one’s evidence); rather, they are reasons for a distinct attitude, suspension, which outweigh the evidential reasons for her given doxastic attitudes. Moreover, say that the fact that Poirot’s belief and his disbelief are incoherent indeed indicates that some of the given evidence (either Miss Marple’s testimony or the first-order evidence on the crime scene) is misleading. In this sense, the incoherence fact is a reason undercutting both pieces of evidence quite independently of a subject’s point in time and of who she is. Say that Colonel Hastings learns of Poirot’s incoherence. The incoherence fact, as evidence that some of the given evidence is misleading, weakens the probatory power of this evidence for Hastings as well, and independently of the stage of his or Poirot’s reasoning processes. But this is not a match for the particular temporal profile and agent-relative nature of my proposed incoherence-based reasons (core feature (4)).

But you might further worry that in 6/49 Lottery, if the incoherence fact really is neither evidence that Lola’s reasoning is faulty nor evidence that her evidence is misleading, then it is also not (or no longer) a reason for her to suspend. For if her reasoning and her evidence are flawless, and Lola knows this, it seems that further inquiry is pointless; at least, she knows she will not be able to resolve the inconsistency. So, the incoherence fact appears to have ceased to be a reason to inquire, and thus to be a reason to suspend. This does not attack my previous point that the reason to suspend in my cases does not arise in virtue of the incoherence fact undercutting Lola’s first-order evidence that p. Rather, it puts pressure on my claim that the reason to suspend, in Lola’s case, is due to the need to reassess one’s epistemic situation (since there is apparently no more pressure of this kind). Does this commit me to the problematic view that Lola is justified to believe each of the jointly inconsistent propositions?

My response is that even if Lola knows that she reasoned correctly and that her evidence is fine, the fact remains that something is not right, given that her beliefs are inconsistent. Although further reassessment is pointless for Lola after several reevaluation cycles, it would be a mistake to settle on beliefs concerning the lottery propositions. Rather, that she is at the moment unable satisfactorily to resolve the inconsistency is a reason for Lola to remain open to a future resolution, even if it would be pointless to actively inquire further.¹⁹ To remain genuinely open, Lola has to suspend; this fact is then also a reason to suspend. So, I am not committed to the problematic view that all of Lola’s lottery beliefs are justified.²⁰

If you are now convinced that my incoherence-based reasons do not function by being (provided by) pertinent evidence, maybe this will lead you to worry whether such reasons are genuinely epistemic, which brings us to the second horn of the dilemma. Say we grant that these reasons to suspend are not evidential and do not affect the probabilities of the believed propositions—doesn't this show that they cannot be right-kind epistemic reasons (e.g., Singh, 2021)? Problem one of the second horn is that incoherence is not a purely epistemic issue, but concerns theoretical and practical deliberation/attitudes equally, and so it—arguably—cannot give rise to genuinely epistemic reasons. This is well illustrated by mixed cases, as in akrasia, where the subject believes she ought to φ, but fails to intend to φ. Her incoherence gives rise to a reason to reassess her attitudes and so to suspend both, and so it would be arbitrary and unmotivated to identify the incoherence fact as an epistemic reason. Identification as epistemic via the reason’s target (a doxastic attitude) breaks down, since both belief and intention are targeted by the incoherence-based reason. Moreover, individuation via the standards of rationality—practical or epistemic—breaks down. For it is simply incoherence that gives rise to reasons to suspend; it does not make sense that incoherence violates epistemic standards of rationality for the belief, but practical standards for the intention (Worsnip, 2021, chap. 1).

Problem two of the second horn is that, since incoherence facts do not bear on individual attitudes, but impact combinations of attitudes, they cannot be reasons for attitudes, but must be reasons to deliberate a particular way. According to Worsnip (2021, chap. 8), they are reasons to structure one’s deliberation a certain way—that a combination of attitudes is incoherent is a reason to treat it as off-limits in deliberation. But if they are not reasons for attitudes, but reasons to act in certain ways in deliberation, they cannot be right-kind epistemic reasons to suspend.

Let me start by addressing problem one. The incoherence-based reason to suspend in my cases relates to standards of correct theoretical reasoning. When a subject realizes that her doxastic attitudes are incoherent, the proper way for her to respond, from the standpoint of theoretical rationality, is to reassess (her evidence for) all involved attitudes, so as to ensure that they, first, cohere with one another, and second, are sufficiently supported by evidence. I suggest we give this a virtue-epistemological spin. Elsewhere, I argue that an exemplarist virtue-theoretic account of the weight of epistemic reasons can explain why the weight of epistemic reasons is attenuated in cases of pragmatic encroachment (Schmidt ms.). Here, my claim is that the intellectually virtuous response to incoherent doxastic attitudes is to suspend, so that one can properly reevaluate one’s epistemic situation. Since incoherence-based reasons relate to standards of theoretical reasoning and to intellectual virtues, they are epistemic reasons.

This contrasts with mixed cases, where incoherent attitudes violate standards of correct moral or practical reasoning, as in the above variant of akrasia. To illustrate, say you believe that all things considered, you ought to participate in the protests against the government’s failure to address the climate crisis. At the same time, you find yourself not intending to participate in these protests. Analogously to the counterexample cases, this is a reason to step back and to reassess your reasons for your belief and your intention. But unlike in the counterexample cases, the reasons you need to reassess are moral reasons, and it is morally virtuous for you to suspend on both attitudes to reevaluate your reasons. The question is: Did you, when you formed your belief, misrepresent your moral reasons—that is, does your belief overstate the strength of your moral reasons to protest? Or alternatively, do your moral reasons, which you accurately represent by way of your belief, fail to motivate you? By engaging in correct moral deliberation, you will either overcome your representative failure (adjust your belief) or fix your motivational problems (adjust your intention). So here, the reason to suspend is a moral reason.

The same holds if we construct the case so that it involves incoherent end intention, means intention, and means-end belief: You intend to put pressure on the government to seriously address the climate crisis. You believe that a necessary means to do so is for you to participate in the local climate protest today. But you do not intend to go to the climate protest. Again, your combination of attitudes fails to meet moral standards of reasoning, and it is a matter of moral virtue to suspend so as to reassess where the mistake lies—whether it is a motivational failing (to respond to the given moral reasons by intending to put pressure on the government, while not intending to go to the protest); or whether your belief fails to adequately represent the practical relevance of going to the protest (maybe it is not a necessary means to putting pressure on the government). In summary, incoherence-based reasons that relate to theoretical standards of reasoning are epistemic, whereas incoherence-based reasons that relate to moral or practical standards of reasoning are moral or practical reasons to suspend.

Turning to problem two, I deny that incoherence facts are exclusively reasons to structure one’s deliberation a certain way; they (typically) also function as reasons to suspend. When Basna, in History vs. Philosophy, realizes that her beliefs are inconsistent, she first and foremost has a reason to suspend on whether any facts are relative, so that she can reassess her epistemic situation. The incoherence of her existing attitudes is a reason for her to suspend, as well as a reason to treat this combination of attitudes as off-limits. Indeed, suspending is the correct way of treating the combination as off-limits. Even if we imagine that Basna (instead of already finding herself with inconsistent beliefs) engages with arguments from historians and philosophers for these beliefs at the same time, and so it is salient that their inconsistency is a reason not to treat accepting both beliefs together as a live option to begin with, it is simultaneously a reason for her to suspend. For she is deliberating on whether the historical facts are relative, or whether no facts are relative; to do so properly, she has to adopt an attitude of being unsettled with respect to these questions, that is, to suspend on them. The same can be argued for many cases of incoherence-based reasons, including ones discussed by Worsnip (2021).²¹

Generally, my discussion makes it clear that incoherence-based reasons exhibit the “earmarks” of right-kind reasons for doxastic attitudes (Schroeder, 2012b, 458; cf. Worsnip, 2018). Not only do they and evidential epistemic reasons share the same target, viz. doxastic attitudes. The cases presented here show that these reasons can be weighed against each other, as when Poirot’s reason to suspend, that he has incoherent doxastic attitudes, outweighs his evidential reasons supporting his belief and disbelief and moves him to suspend on whether v and on whether the evidence indicates that v. The cases further show that subjects are naturally motivated to suspend on the relevant issues by incoherence-based reasons and that such reasons bear on the epistemic rationality of suspension. For instance, subjects like Lola are intuitively epistemically justified to suspend when they realize they have inconsistent attitudes and not justified to uphold any of the involved beliefs.

6 Results

In this paper, I have presented and defended a counterexample to (R-E), the view that all epistemic reasons are provided by evidence. The counterexample from incoherent doxastic attitudes involves incoherence-based epistemic reasons to suspend on a matter which cannot be traced back to evidence concerning that matter, or so I have argued.

My motivation for developing the counterexample has been to put sufficient distance between reasons-first views in epistemology and evidentialism to establish the former as views of independent interest. If I am right, evidence is something, but not the only thing that makes doxastic attitudes epistemically reasonable. Epistemic reasons go above and beyond evidence, and pursuing a reasons-centric epistemology has the potential to allow us to gain insights that are not available to evidentialists. On the one hand, such insights may be won by inquiring into the parallels and differences between epistemic and practical reasons that motivate some philosophers to endorse a reasons-first epistemology (Schroeder, 2021). On the other hand, it is an important avenue of research to clarify how exactly epistemic reasons go beyond evidence—what is the specific epistemic way in which epistemic reasons favor, given that (R-E) is incorrect?

Above, I have argued that incoherence-based reasons have to do with the subject’s appropriate intellectual conduct. One might generalize this idea by claiming that what makes epistemic reasons specifically epistemic is that they guide the subject in appropriately regulating her intellectual conduct, that is, with a view to the truth. Evidence-based reasons are clearly able to play this role, but so are incoherence-based reasons to suspend. Such a proposal can be tied in with an approach that spells out standards of correct theoretical reasoning in virtue-epistemological terms, that is, by way of what would be the intellectually virtuous response to certain facts. I aim to develop such an account in detail on another occasion.