Fittingness first?: Reasons to withhold belief

Abstract

Recent years have seen the rise of fittingness-first views, which take fittingness to be the most basic normative feature, in terms of which other normative features can be explained. This paper poses a serious difficulty for the fittingness-first approach by showing that existing fittingness-first accounts cannot plausibly accommodate an important class of reasons: reasons not to believe a proposition. There are two kinds of reasons not to believe a proposition: considerations that are counterevidence; and considerations that count against believing the proposition without indicating that it is false. I will argue that the fittingness-first accounts have trouble accommodating reasons of the latter kind.

A popular and attractive idea about normativity is that there is a fundamental normative concept or property in terms of which all of the other normative concepts or properties can be analyzed. If this is right, there is a promise of having a unified picture of the normative domain, and metaethics can fruitfully fix attention to a single normative feature and explore its nature. The question is, of course, which normative feature comes first. A popular view is that reasons come first: reasons should be taken as basic and other normative features, such as good, right, rational, ought, and so on, should be built out of reasons.¹ A prominent alternative is the view that good or value comes first.² Yet another, increasingly influential option is the view that fittingness is the most basic normative element. Early proponents of the view include Franz Brentano (1889/2009) and Ewing (1948). More recently, versions of the view have been defended systematically by Timothy Chappell (2012), Conor McHugh and Jonathan Way (2016), and Christopher Howard (2019).

This paper argues against the fittingness-first view by way of arguing against the idea that reasons can be fully accounted for in terms of fittingness. The question whether reasons can be explained in terms of fittingness is distinct from the question whether fittingness comes first: even someone who rejects the idea that fittingness is the fundamental normative unit might consistently hold that reasons can be explained in terms of fittingness. Clearly, however, if some facts about reasons resist explanation in terms of fittingness, it poses a challenge for the fittingness-first view.

The existing fittingness-first accounts are divided into two types. According to the first, reasons are defined as explanations of fittingness. According to the second, reasons are defined as premises of fittingness-preserving reasoning. I shall argue that neither version can plausibly account for an important class of reasons. There are two kinds of reasons not to believe a proposition: considerations that are counterevidence; and considerations that count against believing the proposition without indicating that it is false. All of the existing accounts, however, fail to accommodate reasons of the second kind. This indicates a big lacuna in the fittingness-first program, since each account represents a highly promising and straightforward way of implementing the idea that fittingness comes first.³

The paper proceeds as follows. Section 1 explicates some prominent fittingness-first accounts and their main motivation. Section 2 argues that Howard’s account fails to explain the second type of reasons against believing: either belief has a perspective-independent fittingness condition or it has a perspective-dependent fittingness condition. If the former, Howard’s account cannot adequately account for the second type of reasons. If the latter, Howard’s account over-generates reasons for belief. Section 3 argues that McHugh and Way’s view falls prey to the same dilemma.

1 Fittingness-based accounts of reasons

Fittingness-first theorists, or fittingness-firsters, hold that all other evaluative and normative features, including reasons, can be analyzed in terms of fittingness. This section briefly explains one of the main motivations for the fittingness-first approach and outlines two leading accounts of reasons in terms of fittingness. While each account might have resources to account also for values, I shall focus on the account of reasons, for the sake of simplicity.

1.1 The motivation: Buck-passing and the wrong kind of reasons

One of the main motivations for putting fittingness first comes from the wrong kind of reasons problem, which has proven to be a central problem for the reasons-first approach. Consider the buck-passing analysis of value, according to which a value property (or concept) can be analyzed in terms of reasons to have a pro-attitude towards the bearer of the property.⁴ For example, an object X’s being valuable (or desirable) is a matter of there being sufficient reasons to value (or desire) X. The analysis can be applied to more specific evaluative properties, such as admirability: for X to be admirable is for there to be sufficient reasons to admire X.

The wrong kind of reasons problem for the reasons-firsters is that there could apparently be reasons to, say, admire a person that are irrelevant to the person’s admirability. For instance, if a billionaire promises to pay you half of his fortune for admiring a thoroughly despicable person, this fact apparently provides you with a sufficient reason to admire them, but without showing them to be admirable. Reasons-firsters have two options here: (i) they can distinguish between the right and the wrong kinds of reasons for pro-attitudes and apply the buck passing analysis only to the former; or (ii) they can simply deny that such incentives are reasons for admiration.

Fittingness-firsters find neither option satisfactory. The problem with the first is that it is just too difficult to draw the distinction correctly without reference to values or fittingness. If the right kind of reasons to admire a person were defined as those reasons that make the person admirable, the buck passing analysis would no longer be an informative analysis of admirability. If the right kind of reasons were to be defined as those reasons that bear on the fittingness of admiring the object, the analysis would include a further normative primitive, namely fittingness, which is outright inconsistent with the reasons-first approach. If one appeals to neither values nor fittingness, however, it is difficult to draw the distinction between the right and the wrong kinds of reasons in a way that is both straightforward and extensionally adequate.⁵

The problem with the second option is, first, that it is counterintuitive. There is a clear sense in which the incentive counts in favor of admiring the despicable person. So, the onus is on the reasons-firsters who take this strategy to offer a plausible explanation of why such incentives cannot be reasons, which seems to be no easy task. For example, McHugh and Way (2016: 582) argue that, so long as the concept of a reason is taken as indefinable or primitive, it is extremely difficult to explain why such incentives for attitudes do not count as reasons.⁶ For example, a popular way of denying that incentives are reasons is to impose a response condition on reasons: a reason to φ must be the kind of consideration for which you can φ.⁷ However, it is unclear how reasons-firsters can explain such a strong constraint on reasons. For nothing in the bare notion of counting in favor of a response seems to conceptually entail such a strong constraint.⁸ Reasons-firsters’ appeal to such a constraint would then amount to positing a brute, unexplained normative truth, which is unattractive.⁹

1.2 McHugh and Way’s account

In light of this difficulty for the reasons-first approach, fittingness-firsters suggest that we reverse the order of explanation. Consider, first, McHugh and Way’s account:

McHugh and Way’s account: For that p to be a reason for a response is for that p to be a premise of a good pattern of reasoning from fitting responses to that response. (McHugh & Way 2016: 586)

To wit, reasons are true premises of good reasoning, or the contents of the beliefs one reasons from in good reasoning. Importantly, the goodness of reasoning is not an independent normative feature, since it is eventually understood in terms of fittingness. On this account, a good pattern of reasoning is a pattern of reasoning that preserves fittingness: a reasoning pattern is good just in case the conclusion-attitude is fitting if the premise-attitude(s) is fitting, other things being equal (McHugh & Way, 2016: 588).

This characterization seems to sit well with ordinary cases of reasoning involving belief and intention. Paradigmatically valid patterns of reasoning, such as modus ponens inference, count as good patterns of reasoning: they preserve fittingness, assuming that a belief is fitting if and only if it is true. The instrumental pattern of reasoning, which consists in a transition from intending to E and believing that M-ing is the only way to E towards intending to M, also turns out to be fittingness-preserving, assuming that an intention to φ is fitting if and only if φ-ing is permissible (or choiceworthy). The ‘other things being equal’ clause is intended to cover defeasible reasoning, which is fittingness-preserving on the assumption of normality, such as reasoning from the belief that a reliable expert said that p to believing that p.¹⁰

On McHugh and Way’s account, paradigmatically right kind of reasons for admiration turn out to be genuine reasons for admiration and paradigmatically wrong kind of reasons fail to qualify as such: it is (defeasibly) good reasoning to move from believing, say, that a person is intelligent to admiring them, whereas it is not good reasoning to move from believing that I will get rich by admiring them to admiring them. This seems to make their account extensionally adequate as an account of the right kind of reasons for admiration. The same holds, mutatis mutandis, for reasons for other attitudes.

1.3 Howard’s account

One consequence of McHugh and Way’s account is that there are no wrong kind of reasons for attitudes, strictly speaking. A purely pragmatic reason for admiring X is not a premise from which you can reason well towards admiring X. It is a reason for some other response, such as intending to admire X, or believing that it is desirable to admire X, etc. One might find the inability to classify the wrong kind of reasons as genuine reasons to be an unattractive feature of McHugh and Way’s account and opt for an analysis that allows both kinds of reasons to be reasons. So, consider:

Howard’s account: For p to be a reason to φ is for p to explain either why it is fitting to φ or why it is fitting to want to φ.¹¹ (Howard 2019: 239)

This straightforwardly makes sense of the intuitive distinction between the right and the wrong kinds of reasons: reasons that explain why it is fitting to φ, or fit-related reasons, correspond to the former; and reasons that explain why it is fitting to want to φ, or value-related reasons, correspond to the latter. For example, facts that explain why it is fitting to admire X are reasons to admire X in virtue of satisfying the first condition, and mere incentives for admiring X count as reasons in virtue of satisfying the second condition, that is, by explaining why it is fitting to want to admire X. In fact, the second condition of Howard’s account also serves to define good in terms of fittingness: for p to be a respect in which x is good is for p to explain why it would be fitting to want x (Howard, 2019: 229).

It might thus seem that fittingness is the fundamental normative unit that can neatly account for reasons and values, which might in turn explain further normative features like ought and rational. In what follows, I shall argue that it cannot.

2 Reasons not to believe: problems for Howard’s account

2.1 Reasons against believing: the problem

A standard assumption in the literature on epistemic reasons is that the right kind of reasons to believe a proposition are evidential considerations. When you have evidence for p, it gives you a reason to believe p. What about reasons against believing p? Some reasons not to believe p are simply counterevidence, that is, reasons to disbelieve p (by which I simply mean believing not-p). I will call them Type-1 reasons against believing p. But there also seem to be reasons not to believe p that are not evidence for not-p.

First, there are facts about evidence, which are neither evidence for p nor evidence for not-p. For example, if you have no evidence whatsoever that bears on p, then this fact plausibly gives you a reason not to believe p. Similarly, if your evidence is equally balanced between p and not-p, this fact plausibly gives you a reason not to believe p. Such facts, however, do not evidentially support p or not-p in themselves.¹²

Second, there are undercutting defeaters, which undermine the supporting relation between the evidence and the target proposition, rather than indicating the falsity of the proposition.¹³ Suppose that you enter an empty room and see the wall in front of you, and justifiably believe that the wall is red, but immediately come to learn that the room is illuminated by a red light. This fact itself is not evidence that the wall is not red, but intuitively gives you, in this situation, a reason not to believe that the wall is red and affect what the balance of your reasons supports.¹⁴

Finally, there is higher-order evidence, a consideration which bears on the quality or the strength of your evidence, or your capacity to assess evidence. Suppose that you solve a simple logic puzzle and become confident that P is the answer, but then you are told that the coffee you sipped before solving the puzzle was laced with a drug which is known to degrade people’s performance in this type of task.¹⁵ It is again intuitive that this fact about the drug, without being evidence that P is not the correct answer, gives you at least some reason not to believe that it is correct.

I submit that an adequate account of reasons should, other things being equal, accommodate the intuitive judgment that such considerations are reasons not to believe a proposition, despite not being evidence that the proposition is false.¹⁶ In what follows, I shall call such reasons Type-2 reasons against belief. I will consider Howard’s account and McHugh and Way’s account in turn and show that neither has a satisfactory account of Type-2 reasons: they cannot explain Type-2 reasons without over-generating reasons.

2.2 Problems for perspective-independent solutions

Recall that on Howard’s account, Type-2 reasons should turn out to be either fit-related reasons or value-related reasons. But Type-2 reasons are plausibly fit-related, rather than value-related reasons. The fact that I have no evidence regarding whether I have an even number of hairs has nothing to do with the undesirability of believing that I have an even number of hairs, and hence does nothing to explain why it is (un)fitting to want to believe that I have an even number of hairs. If this reason is a fit-related reason not to believe, however, Howard’s account should offer an explanation of why.

One natural suggestion is that a consideration is a fit-related reason not to believe p just in case it explains why it is unfitting to believe p. If so, the fact about my lack of evidence would turn out to be a reason not to believe that I have an even number of hairs if it explained why it is unfitting to believe that I have an even number of hairs. However, it is unclear how this fact could do so, if one assumes the following:

Truth

It is fitting to believe p if and only if p is true.¹⁷

Assuming plausibly that Truth holds (if it holds) as a matter of conceptual necessity, a consideration that explains why it is (not) fitting to believe p also explains why p is (not) true. This already saddles Howard’s account with significant problems that are independent of the task of accounting for Type-2 reasons. First, it is unclear how all pieces of evidence for p can count as a reason on this account. For, in many cases, evidence does not explain why, but rather indicates that, a proposition is true. For example, Smith’s fingerprint on the murder weapon does not explain why it is true that Smith is the murderer: if at all, it is the latter that explains the former. Second, this account seems to imply that there are no reasons to believe a false proposition: if p is false, it follows from Truth that it is not fitting to believe p, so nothing can explain why it is fitting to believe p.

If Howard takes a reason for a response to be evidence of its fittingness, rather than an explanation of it, then his account might be able to overcome such problems.¹⁸ But Type-2 reasons still remain a problem, whether reasons are explanations or evidence. The fact that I have no evidence about the number of my hairs does not explain why I do not have an even number of hairs. It only explains why, given my total evidence, I am not positioned to know that (or it is insufficiently probable that) I have an even number of hairs. Likewise, this fact is not evidence that I do not have an even number of hairs: it is just evidence that I am not positioned to know that I have an even number of hairs.

Another strategy that Howard might take is to hold that Type-2 reasons are not reasons for the absence (or lack) of a belief, but rather reasons for withholding (or suspending) belief, understood as a genuine doxastic attitude.¹⁹ Assuming plausibly that believing p and withholding belief about p are alternative doxastic responses, and that reasons for a response are reasons against its alternatives, a consideration might count against believing p by being a reason for an alternative, namely, withholding belief. On this basis, Howard might hold that Type-2 reasons not to believe p just are considerations that explain why it is fitting to withhold belief about p.

The problem with this move is that it seems impossible to specify the conditions under which withholding belief about p is fitting in a way that is consistent with both (i) Truth and (ii) an independently compelling principle about the fittingness of belief and withholding. Plausibly, whether it is fitting for you to withhold belief about p depends on your epistemic position. For example, if neither p nor not-p is probable given your available evidence, it is fitting for you to withhold about p, regardless of whether p is true. Assuming Truth, this means that there can be situations in which it is both fitting to believe p and fitting to withhold belief about p. But this contradicts the following, highly plausible principle about when belief and withholding are fitting:

Belief-withholding link

If it is fitting for you to believe p, then it is not fitting for you to withhold belief about p.²⁰

The idea that belief is fitting if and only if it is true derives its plausibility from the idea that having a fitting attitude is a matter of getting things right. And whenever p is true, only believing p is a way of getting things right and neither withholding nor disbelieving gets things right. For a helpful analogy, imagine a horserace which horse A has won. It is only those who bet on A who can be said to have gotten things right: those who decided not to make a bet or those who bet on another horse have, each in their own way, failed to get things right. If so, Belief-Withholding Link is true. This prevents the defenders of Howard’s account from assigning a plausible fittingness condition to the attitude of withholding belief, which is the very first step of accommodating Type-2 reasons as reasons to withhold belief.

One might attempt to block this argument by rejecting Belief-Withholding Link. For example, Rosa (2020: 19–20) distinguishes between ex ante correctness and ex post correctness of having a doxastic attitude and argues that even if withholding-belief-about-p-and-believing-p is always ex post incorrect, there are possibilities in which it is both ex ante correct to believe p and ex ante correct to withhold belief about p. His reasoning is this: if p is true and it is possible for the subject to correctly believe p, then it is ex ante correct for the subject to believe p. But if, at the same time, the subject’s situation (understood as the total information encoded by her cognitive system) leaves it open whether p is true and it is possible for her to correctly withhold belief about p, it is also ex ante correct to withhold. This might seem to entail that it is ex ante correct to withhold-belief-about-p-and-believe-p, which is highly implausible. Rosa denies this, however, on the grounds that (ex ante) correctness does not agglomerate over conjunction, just as some normative status such as permissibility does not agglomerate over conjunction: you might be permitted to text your friend and permitted to drive, without being permitted to text-and-drive.

The problem is that adapting this move to the case of fittingness is not plausible.²¹ First, the analogy with permissibility is dubious and so it is unclear that fittingness does not agglomerate over conjunction. To say that a response is fitting is not just to say that it is not forbidden: it is to say that there is a suitable match between the response and its object, a natural metaphor for which is, as Berker (forthcoming) puts it, a “key fitting into a lock”. For example, when one says of a shameful action that it is fitting for the agent to be ashamed of it (or that it merits shame), one is intuitively not just saying that the agent is permitted to be ashamed at their conduct.²² But if fittingness consists in such a match, it is hard to see why it fails to agglomerate: if an attitude (say, admiring X) matches its object and another attitude (say, fearing Y) matches its object, what prevents admiring-X-and-fearing-Y from being fitting? (Each key has fit into its lock!) Secondly, it is unclear why fittingness is supposed to be dependent on one’s perspective in the case of withholding belief but not in the case of belief. The alleged asymmetry would remain ad hoc absent independent motivation for it, which seems difficult to find. Plausibly, the idea that it is fitting to withhold belief about p when one’s situation leaves it open whether p is true owes its plausibility to taking fittingness as (rational) appropriateness or reasonableness. But once fittingness is taken as such, believing p would not be fitting in the same situation. For it is neither appropriate nor reasonable to believe p when one’s situation leaves p’s truth open.

2.3 Problems for perspective-dependent solutions

However, Truth is not forced upon fittingness-firsters. For the fittingness of your belief might well depend on your epistemic position or perspective. Indeed, there is a clear sense in which it is inappropriate to believe a proposition that is improbable given one’s evidence, even when it is true. And one might think that to call a belief inappropriate in this way just is to call it unfitting. This opens door to alternative ways of specifying the fittingness condition of belief, such as:

Evidence

It is fitting (for you) to believe p if and only if p is sufficiently supported by your available evidence.

Knowledge

It is fitting (for you) to believe p if and only if you are in a position to know p.²³

Both take the fittingness of your belief to depend on your epistemic position, in requiring that your belief be supported by available evidence. One difference is that whereas Evidence does not require the truth of a belief as a necessary condition for its fittingness, Knowledge does require the truth, since being in a position to know p entails the truth of p (cf. Lord, 2018b).

Howard’s account, coupled with either Evidence or Knowledge, can neatly account for Type-2 reasons. The fact that I have no evidence about the number of my hairs explains why it is not sufficiently likely that I have an even number of hairs, given my evidence. It can also account for undercutting defeaters: the fact that the room I entered is illuminated by a red light explains why it is not sufficiently likely that the wall is red, given my evidence.

The problem is that Howard’s account cannot combine with either Evidence or Knowledge without over-generating reasons. To see this, recall that anything that explains why it is fitting to believe p is a (fit-related) reason to believe p, on Howard’s account. Together with Evidence, it entails that anything that explains why p is sufficiently supported by your available evidence is a reason to believe p. But this seems false. Suppose that Anne has decisive available evidence (or is in a position to know) that anthropogenic climate change is happening. Let us also assume the following: what partly explains why Anne is in the epistemic position she is in is that she has never met any climate change skeptics and her only sources were those who believe in climate change. This means, on Howard’s account (coupled with either Evidence or Knowledge), that the fact that Anne’s only sources were those who believe in climate change is a reason for her to believe that climate change is happening, which seems to be a wrong result.

One natural response is that my objection features an explanation of the wrong kind and Howard’s account might avoid this result if the relevant explanations are restricted to explanations of the right kind. This response has problems, however. The first is a dialectical one: one major advantage claimed for fittingness-first views is that they have an easy time dealing with the wrong kind of reasons problem. But if fittingness-firsters had to offer a distinction between explanations of the right and the wrong kinds, it would seem that they have simply swept the real problem under the rug. The second problem is that it is difficult to draw a principled distinction in a way that would avoid the problem at hand. One natural move for Howard to make, for example, is to hold that only constitutive (as opposed to causal) explanations are relevant. But this restriction hardly solves the problem. Suppose that Howard’s account is coupled with Evidence. The fact that the body of evidence E available to Anne indicates that climate change is happening is partly constituted by the fact that E is available to Anne, which is a matter of her having epistemic access to E. Thus, the fact that Anne has access to E partly (and constitutively) explains why Anne’s evidence indicates that climate change is real. Howard’s account implies that this fact is a reason for her to believe that climate change is real, which does not seem right: it is her evidence itself that is such a reason, not the fact that she has access to it.

Combining Howard’s account with Knowledge fares no better. Consider the following facts: (a) Alice is an expert on climate change; (b) Alice has gathered all the relevant data; (c) Alice can competently evaluate the data; (d) Alice has reasoned correctly from the data that climate change is happening; (e) It is true that climate change is happening. Plausibly, each of these facts partly and constitutively explains why Alice is in a position to know that climate change is happening. But Howard’s account-cum-Knowledge entails that each consideration is a reason for Alice to believe that climate change is real. Again, however, this is implausible.²⁴ For example, the fact that Alice is an expert does not, on its own, seem to be a reason for her to believe that climate change is happening.²⁵

One might object that a consideration’s being a partial explanation of the fittingness of a response is not sufficient to make it a reason for the response. Instead, one might hold that it has to be a complete explanation of its fittingness. In Alice’s case, for example, the complete explanation of why she knows that climate change is happening would be given by the conjunction of (a)-(e), and none of the individual facts would provide such an explanation. If so, one might insist, Howard’s account-cum-Knowledge can avoid this result.

But the completeness requirement would set too high a bar for a consideration’s being a reason for belief. Suppose that a fairly reliable weather forecast predicts that it will snow today, but you also feel that the temperature today is too high for it to snow. It is highly intuitive that the weather forecast gives you some reason to believe that it will snow today, and the felt temperature gives you some reason not to believe it. But neither completely explains why it will (not) snow or why you are (not) in a position to know that it will snow, and so neither would qualify as a reason (not) to believe that it will snow, given the completeness requirement. Indeed, the requirement seems to imply that no consideration counts as a reason unless it is also decisive. But there is no reason to impose such a demanding condition on reasons.

To sum up the objection, Howard’s account faces a dilemma. Either the fittingness of your belief depends on your epistemic position or not. If the former, then Truth holds, and Howard’s account cannot adequately explain why Type-2 reasons are reasons against believing. If the latter, either Evidence or Knowledge is true, and Howard’s account over-generates reasons for belief.

3 Reasons not to believe: problems for McHugh and Way’s account

3.1 The problem, redux

Let us now consider McHugh and Way’s account, according to which a reason is a premise of good reasoning from fitting beliefs. The problem of Type-2 reasons might also afflict this account. If a consideration C is a reason for you not to believe p (or to withhold belief about p), such an account is committed to the view that if your belief in C is fitting, not believing p (or withholding belief about p) is fitting, other things being equal. Assuming Truth, however, one might find it unclear how such a conclusion-attitude can be fitting at all, on the following grounds: (i) either p is true or not; (ii) if the former, it is fitting to believe p; (iii) if the latter, it is fitting to disbelieve p. Thus, assuming Belief-Withholding Link (from 2.2), not believing p (or withholding belief about p) can never be fitting.²⁶ By McHugh and Way’s lights, this means that a pattern of reasoning which concludes with non- or withheld belief cannot be (non-trivially) fittingness-preserving, which means that it cannot be a (non-trivially) good pattern, which seems problematic.

3.2 Problems for perspective-independent solutions

An obvious response to the problem is to hold that reasoning towards the absence of a belief is good just in case its absence is fitting if the premise-responses are fitting. One problem with this proposal is that, as McHugh and Way (forthcoming) note, it is unclear that the absence of an attitude can be fitting: fittingness consists in a suitable match between an attitude and its object, but an absence does not have any object. A deeper problem is that, even if it is granted that the absence of a belief can be fitting, the most natural view given Truth is that it is fitting just in case the belief is unfitting (that is, false). But then McHugh and Way’s account has difficulty accommodating Type-2 reasons. For example, it is not good reasoning to move from the belief that I have no evidence about p towards believing that p is false.

McHugh (2014: 184) himself considers an alternative view, which is based on the following:

Assumption

Transitioning from a premise-response(s) to the absence of a response is a good pattern of reasoning just in case it is not a good pattern of reasoning to move from the premise-response(s) to the response.

From Assumption and McHugh and Way’s account it follows that if it is not a good pattern of reasoning to move from a premise-response(s) to believing p, the premise is a reason not to believe p. Since it is not a good pattern of reasoning to move from one’s belief in Type-2 reasons to believing the target proposition, it follows that they are reasons not to believe. So far so good.

The problem, which McHugh himself notices, is that this view over-generates reasons. Suppose I reason from the belief that grass is green towards the belief that Joe Biden is the president of the United States. This is not a good pattern of reasoning. The view in question entails that the fact that grass is green is a reason not to believe that Biden is the president. Intuitively, however, this is neither a reason to believe, nor a reason not to believe, that Biden is the president. Moreover, the point generalizes to reasons not to have attitudes of other kinds. For on the proposed view, the fact that grass is green is a reason not to intend to work, a reason not to admire Nelson Mandela, a reason not to fear a tiger, and so on.

McHugh is willing to bite the bullet on this score. One reason he cites to soften the bullet is that, if all the evidence I have available is that grass is green, then I ought not believe that Biden is the president. However, it seems that this is true not because there is a reason for me not to believe, but simply because you do not have any positive reason to believe, and it is a distinctive fact about belief that you ought not believe when you do not have any positive reason to believe. For example, McHugh’s proposal has an implausible consequence regarding reasons for intention. Take an arbitrary action which there is intuitively no reason for or against, such as rubbing your nose for a second. Assuming plausibly that it is permissible for you to (intend to) φ if there is no reason for you not to φ, it is permissible for you to intend to rub. Suppose now that the only thing you know is that grass is green. Intuitively, this fact does not affect the deontic status of rubbing. On McHugh’s proposal, however, the fact that grass is green is a reason for you not to intend to rub, which means that intending to rub is no longer permissible: your reasons against intending to rub are now weightier than your reasons for intending to rub.

3.3 Problems for perspective-dependent solutions

Here again, one natural option is to reject Truth and opt for either Evidence or Knowledge. It does not seem open for McHugh and Way to accept Evidence, since their account of good reasoning is built on the assumption that a belief is fitting only if it is true. Still, McHugh and Way’s account is consistent with Knowledge, on the basis of which one can plausibly specify the conditions under which withholding belief is fitting: it is fitting for you to withhold belief about p if and only if you are not positioned to know whether or not p. If one combines this fittingness condition for withholding with McHugh and Way’s account, then one has a natural explanation of how you can reason well from beliefs in Type-2 reasons to withholding belief about the target proposition: if my belief that I have no evidence about the number of my hairs is fitting, withholding belief about whether I have an even number of hairs is fitting.

However, McHugh and Way’s account also has problematic consequences, when combined with Knowledge. My first argument turns on the KK principle, which states that if you are positioned to know p, you are positioned to know that you are positioned to know p. While KK is not so popular, notably due to Williamson’s (2000) influential argument against it, there are considerable arguments in its favor, so let us assume it for the sake of argument.²⁷ Recall that McHugh and Way’s account allows for defeasibly good reasoning patterns (1.2), which preserves the fittingness of the premise-attitudes under normal conditions. Given KK, the following ceteris paribus principles seem undeniable:

Knowledge-evidence link

Normally, if you are in a position to know p, you are in a position to know that you have sufficient evidence for p.

Knowledge-competence link

Normally, if you are in a position to know p, you are in a position to know that you are competent to tell whether p.

For being in a position to know p requires having both sufficient evidence for p and relevant competence, setting aside some exceptional cases where it seems that one can know without further evidence, such as cases of proprioceptive knowledge. But it follows from Knowledge and Knowledge-Evidence Link that if believing p is fitting, then believing that you have sufficient evidence for p is fitting, other things being equal. It also follows from Knowledge and Knowledge-Competence Link that if believing p is fitting, then believing that you are competent to tell whether p is fitting, other things being equal. But this means that the following patterns of reasoning are (defeasibly) fittingness-preserving, and so turn out to be good patterns of reasoning on McHugh and Way’s account:

Pattern 1

P; therefore, I have sufficient evidence for p;

Pattern 2

P; therefore, I am competent enough to tell whether p.

But if Patterns 1 and 2 are good, McHugh and Way’s account entails that the fact that p is a reason to believe that you have sufficient evidence for p, and is also a reason to believe that you are competent enough to tell whether p. This is highly problematic. In general, the mere fact that p cannot plausibly be a reason for anyone to believe that they are in a good epistemic position regarding it, since it indicates nothing about their evidence or competence.

Since this objection depends on KK, McHugh and Way could reject KK in order to avoid these consequences. But one might reasonably fault an account of reasons that is invariably committed to the denial of KK. Be that as it may, I will now present an argument that does not rely on KK. Consider the following, which seems clearly true:

• (1) If I fully grasp the relevant concepts (e.g. <equiangular>, <equilateral > and < triangle>), I am in a position to know that all equiangular triangles are equilateral triangles.

Now, (2) follows from (1) and the factivity of being in a position to know:

• (2) If I am in a position to know that I grasp the relevant concepts, I am in a position to know that all equiangular triangles are equilateral.

Given Knowledge, it follows from (2) that:

• (3) If it is fitting to believe that I grasp the relevant concepts, it is fitting to believe that all equiangular triangles are equilateral.

McHugh and Way’s account and (3) entail that the following is a good reasoning pattern:

Pattern 3

I grasp the relevant concepts; therefore, all equiangular triangles are equilateral.

On McHugh and Way’s account, this means that the fact that I grasp the relevant concepts is a reason for me to believe that all equiangular triangles are equilateral. This, again, is an implausible result. If I am asked what reasons I have to believe that equiangular triangles are equilateral, facts about equilaterality and equiangularity that together entail the truth of the belief would be aptly cited as reasons, but facts about my concept possession would not.²⁸ Note that this is not an isolated example: any piece of a priori knowledge (e.g. Bayes’ theorem) based on conceptual competence (e.g. probability concepts) can lead to the same problem.

One response open to McHugh and Way is to invoke their distinction between competent and incompetent reasoning (McHugh & Way, 2016, 2018). On their view, one can reason in conformity with a fittingness-preserving pattern incompetently: one’s reasoning might not manifest sensitivity to the fact that it is fittingness-preserving. An example is a person who concludes that 79 is a prime number from any belief they happen to have. Such a pattern is (assuming Truth) fittingness-preserving simply because it is necessarily true that 79 is prime, but the person is not sensitive to the fittingness-preserving character of the transition. Still, it seems unclear whether McHugh and Way can plausibly count Patterns 1, 2, and 3 as the ones that are not competently followable. There seems to be no reason to think that someone following Pattern 1, 2, or 3 must be conforming to this pattern only accidentally, without being sensitive to their fittingness-preserving character.²⁹ Imagine a person who is convinced that believing p is fitting if and only if they are in a position to know p, and is also convinced that Knowledge-Evidence Link is true. Suppose that this person carefully, correctly works out the implications of these claims and concludes that they have good evidence for p from their belief in p. Intuitively, they have been self-consciously sensitive to the fittingness-preserving character of reasoning. Thus, the problem of overgeneration is yet to be addressed.

4 Conclusion

In this paper, I argued that two influential and well-developed fittingness-based accounts of reasons fail to offer a plausible account of reasons against believing that are not counter-evidence: either they fail to account for such reasons or they over-generate reasons for belief.³⁰ Of course, I have not considered every fittingness-first account of reasons in the logical space and so even the success of my arguments would not disprove the idea that fittingness is the fundamental normative property. Still, my arguments do show that an extensionally adequate fittingness-first account would have to be markedly different from any of the standard accounts in the literature, and that there is a good reason not to believe that fittingness comes first until such an account appears.³¹