1. The alpine climber

Perhaps the most well-worn example of doxastic efficacy is the alpine climber case.Footnote ⁴ You will survive if you leap with determination, but if you hesitate then you are lost. James writes:

James writes that this “belongs to an enormous class” of cases in which it is wise to “believe what is in the line of your needs, for only by such belief is the need fulfilled” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 53–54).

A&T argue that it is permissible for you to believe in such a case but that it is not a counter-example to staunch evidentialism. If you are a reflective mountain climber who knows that your belief will have doxastic efficacy in this case, then (they claim) your confident belief that you will survive does not lack sufficient evidence. Instead, they argue, your belief and the fact that it is efficacious are themselves evidence for the belief (Aikin and Talisse Reference Aikin and Talisse2018: 66). Call this the extra evidence view. Versions of this view are also advanced by Wood (Reference Wood2008: 10) and Misak (Reference Misak2013: 63).Footnote ⁵

In the case as James imagines it, you will survive if and only if you believe.Footnote ⁶ Allowing for degrees of belief modeled in the usual way, we can generalize the case and say that the objective probability that you will survive is equal to your personal credence that you will survive. Formally, Pr(S) = Cr(S).Footnote ⁷

Pr(S) = Cr(S) together with Cr(S) = 1 would provide evidence for S to a third-person observer. They entail that Pr(S) = 1. Yet this move is not available to a first-person alpine climber.

There is something unstable about taking your own credence that S as evidence for S. Note that Pr(S) = Cr(S) may hold even if the belief has no causal influence on S. If you wisely try to assign Cr(S) so that it matches your best estimate of Pr(S) and your estimate is correct, then the equality will hold. Moreover, you will believe that the equality holds just because you believe that your estimate is correct. Allowing the value of Cr(S) to be an extra piece of evidence threatens to provide you with such additional evidence for every belief once you reflectively notice that you believe it.

Taking your belief that S as evidence that S is odd for the further reason that it enters the scene too late. There is a moment on the mountain in which you believe S or you do not. You confidently jump or you die. Taking that moment in slow motion, you reflectively consider what your credence ought to be. The evidence prior to that moment does not settle matters, even for a third party. Knowing that Pr(S) = Cr(S), a detached observer may conclude that you will form a true belief in that moment. If you are confident (Cr(S) = 1) then you will survive. If you are disbelieving (Cr(S) = 0) then you will die. In either case, your belief will accurately reflect the facts. Just as the detached observer cannot say how likely you are to live, there is no rule of evidence which settles what you ought to believe. In the generalized case, you can form any credence Cr(S) whatsoever and have an accurate belief. It is better for you to believe than not to, but the evidence simply does not exist until you have formed a belief one way or the other. Making a choice is different than responding to evidence, but A&T elide one into the other in accepting the alpine climber case.

One might defend the extra evidence view by noting that knowing that your situation is one of doxastic efficacy is stronger than knowing Pr(S) = Cr(S). The extra evidence, doxastic efficacy, is that the value of Cr(S) causally determines the value of Pr(S). However, this alone still allows for Pr(S) to take on any value. The extra evidence view requires that your belief itself enter into evidence for that belief itself – that is, that the value of Cr(S) play a role in arriving at a value for Cr(S). The view runs afoul of an irreducible difference between choice and evidence.

It is worth noting, moreover, that the intuitive force of the alpine climber case does not require the precise equality Pr(S) = Cr(S). Even though James claims to be considering cases where “our faith beforehand in an uncertified result is the only thing that makes the result come true” (James Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 53), belief need not be the only thing at work. Even if you believe, perhaps a freak gust of wind will nudge you away from your target and to your doom. If we imagine that there is a 10% chance of some interference causing your jump to fail, then the equation becomes Pr(S) = 0.9 × Cr(S). From this it follows that your credence Cr(S) can only correspond precisely to the objective probability Pr(S) if Cr(S) = 0. That is, they are equal if you believe without any doubt that you will die. A&T's reply to the case supposes that you are reflective and driven to have your beliefs match the facts. In this revised version, their principle requires that you believe you will die and that you do in fact die. Contrariwise, James would insist that you are free to believe with whatever gusto you can muster.Footnote ⁸ Belief is an effective strategy with a 90% chance of success, and a 90% chance of survival is pretty good. It is better for you to believe and probably survive, even though it decouples your credence from the objective probability.

To sum up: A&T adopt the extra evidence view, claiming that a suitably reflective alpine climber may be seen as responding to evidence suitably construed. The view fails because (a) the construal of evidence required is unstable and (b) a variant of the case drives a wedge between James’ response to the case and extra evidence views.

2. Personal relations

In The Will to Believe, James offers examples of “personal relations” – friendship and love. Considering the question “Do you like me or not?” he writes, “Whether you do or not depends, in countless instances, on whether I meet you half-way, am willing to assume that you must like me, and show you trust and expectation. The previous faith on my part in your liking's existence is in such cases what makes your liking come” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 29). A&T object that this “sounds more than a little slimy” and reflects a “darker side” (Reference Aikin and Talisse2018: 70, 71). Indeed, a stalker might well defend invasions of privacy and threatening intrusions on the grounds that the target of their stalking simply must love them. A&T are right to note, “Sometimes the reason why one doesn't have evidence that another is romantically interested is because the other is not interested” (Reference Aikin and Talisse2018: 71, emphasis in original). However, recall the dialectic. James is merely aiming to show that one may sometimes permissibly believe without much if any evidence. It shows too little to note just that there are some cases in which one ought not.

Forming and maintaining human relationships is complicated. In the growth of a romantic relationship, it may be that one partner says ‘I love you’ before the second is prepared to say it back. The first partner may, at this stage of the relationship, believe that the two love each other. In some cases, the second eventually says ‘I love you, too’, and they live happily ever after. Perhaps the relationship would fall apart rather than continuing if the first partner did not believe in their love through a period in which the evidence was ambiguous. Of course, it may be that the second partner subsequently shows through their actions and words that they do not love the first. There is a point at which the evidence is sufficiently decisive that the first partner would be obtuse to continue believing that the second loves them.

So there obviously are cases (like a stalker) in which belief in the utter absence of evidence would be inappropriate and others (like a neglected or abused partner) in which counter-evidence should overwhelm belief. All that James needs, though, is that there are some situations in which the evidence alone is ambiguous and insufficient but in which one may nevertheless permissibly believe. The formation of romantic relationships is at least sometimes like that.

James thinks the point applies to non-romantic friendship, too. One may develop the point in this way: Being friends with someone requires trusting them. Insofar as one only depends upon them as far as evidence shows them to be reliable, one is merely relying on them and not trusting at all. Friendship requires more. As Alexander Nehamas writes, “My friends are people from whom … I don't yet know exactly what I want to get, because I trust them enough to let them influence what I believe and what I desire in ways I would not be able to do … on my own” (Nehemas Reference Nehamas2007: 58). On Nehamas’ account, being friends with someone involves the belief that there is indefinite pleasure and benefit to be gained by interacting with them – as he puts it, recognizing a beauty in them which is not some apprehended pleasure but is instead the promise of happiness. Importantly, for Nehamas, beauty is not a property that inheres in the object. Instead, it is a trust in indefinite and uncertain future goods. It “points to the future” in an open-ended way (Reference Nehamas2007: 63). Thus, friendship and beauty involve beliefs about what will happen that outrun our evidence.

A&T might respond by rejecting Nehamas’ theory of friendship, but a similar point can be made from more quotidian considerations.Footnote ⁹ Getting to know someone and becoming friends with them often requires a commitment of time. Although “one most certainly can be friendly to another person without holding the belief that the other is one's friend” (Reference Aikin and Talisse2018: 69), as A&T maintain, there is effort involved in spending time with someone and making overtures of friendship. Expending the effort with one person precludes expending it with someone else. Even if one cultivates many friendships, one cannot be friends with everyone. So, why this person or these people rather than others? It would be possible, I suppose, to begin either with people whom evidence suggests are promising as friends or to make the choice randomly. It is more common, I think, to act on a sense that another person holds the promise of being a friend. This sense is suggestive, but it is not sufficient evidence. A third-person observer or someone else in your situation need not see the same promise. Thus, one often believes ‘This person has more promise as a possible friend than other people do’ in the absence of sufficient evidence. And that person only has the potential to become your friend if you do make overtures to them. So it seems to be an example of doxastic efficacy. One's life would be impoverished if one refused to believe in any such case.

To defend A&T, one may object that the belief which one forms in the absence of evidence is not that a person is already one's friend. Let us introduce the locution ‘X is friend-apt’ as shorthand for ‘X has more promise as a possible friend than other people do’. Grant for the sake of argument that friend-aptness is a complex disposition that a person possesses regardless of whether they actually do become one's friend or not. So ‘X is friend-apt’ is distinct from ‘X will be one's friend’. If one acts to become friends with someone whom one believes to be friend-apt, then one is more likely to trigger the disposition and actually become friends with them. Doxastic efficacy of a belief that P, however, requires that the belief positively influence the chance that P obtains. The belief (that they are friend apt) and the outcome (that they actually become your friend) are distinct, and so it is not an example of doxastic efficacy.

The case is parallel to others which obviously do not involve doxastic efficacy. For example, suppose you endeavor to dissolve a lump of gold. You believe that gold has a particular disposition – that it is soluble in aqua regia. Your belief allows you to succeed in dissolving the gold, but gold is soluble in aqua regia whether or not you believe that it is. So, too, for someone being friend-apt on the assumption that friend-aptness is a disposition.

Even granting all that, the objection does not favor A&T over James. The example serves James’ purpose but lacks doxastic efficacy. Our rewarding personal relations may require that we sometimes believe in the friend-aptness of others in the absence of compelling evidence. James’ intuition, which I share, is that this suffices to show that such a belief can be permissible. Note that the attempt to become friends with someone is, among other things, a test of how friend-apt they are. So evidence eventually accumulates to settle the matter one way or the other. We will look at other cases that have this structure in the next section. For now, the upshot is that the early belief in friend-aptness serves as a counterexample to the staunch evidentialist claim that one ought never form beliefs in the absence of sufficient evidence.

To sum up: (a) Some romantic entanglements allow one party to believe that both are in love before there is sufficient evidence for that belief, at least for a while. (b) At least one philosophical account of friendship entails that belief in the absence of sufficient evidence is constitutive of friendship. (c) Apportioning effort to befriend somebody rather than somebody else reflects an assessment of friend-aptness which involves belief in the absence of sufficient evidence.

3. Scientific investigation

Having dealt with the usual secular examples, let's consider another.

Alexander Klein (Reference Klein, Baghramian and Marchetti2018) offers a case of scientific investigation that seems to fit the Jamesian will-to-believe pattern. Note that Klein is writing about the contrast between James’ epistemology and Quine's, so he does not draw out all the lessons from the case that I draw out below.

In the early 1980s, it was believed that peptic ulcer disease (PUD) was caused by excess stomach acid. Barry Marshall and his collaborators found a correlation between the presence of the bacteria Heliobacter pylori (H. pylori) and PUD. However, it was unclear whether the bacteria caused the ulcers or was merely present because the stomach was already compromised. It was also known that almost all cats are infected with Heliobacter, but nevertheless PUD is not pandemic among cats. Experiments to infect piglets with H. pylori showed no effect. Marshall decided that further experiment was required, so he ingested a vial of H. pylori. He got sick, and his illness along with biopsies proved that H. pylori is a pathogen. Further research reinforced the connection, and Marshall won the 2005 Nobel Prize in Physiology or Medicine (along with his collaborator Robin Warren).

Klein argues that Marshall's self-experiment shows that he believed, in advance of sufficient evidence, that H. pylori causes PUD. Klein writes, “James … would have to say that Marshall faced a choice of belief” and that Marshall's “willingness to swallow a vial of bacteria that, at the time, had not been widely studied reflects stunning confidence in his attending belief that he understood the causal pathways of H. pylori” (Reference Klein, Baghramian and Marchetti2018: 238).

Matters before Marshall's self-experiment were equivocal. The evidence could be interpreted to favor Marshall's belief, but it need not be. He saw connections with earlier work and other unexplained findings, but other scientists were unconvinced. So Marshall's case, as Klein describes it, is one in which Marshall believed in the absence of sufficient evidence. Since epistemic strictures should not block meritorious, prize-winning science or condemn it as irrational, we should say that Marshall permissibly believed in advance of the evidence.

Klein writes, “Particularly for an early-career researcher, choices about what experimental program to pursue (and thus about what hypotheses one should believe) are inevitably tied up with one's desire and fears about one's future, about one's ability to provide for one's family, about one's own prospects for an interesting and fruitful career, and so on” (Reference Klein, Baghramian and Marchetti2018: 239). Philosophers of science, especially following Kuhn, have often noted this kind of personal investment in theories.Footnote ¹⁰

Note that James himself does not offer this kind of example. On the contrary, he supposes that scientists should exercise caution and remain agnostic until sufficient evidence is collected. He writes, “A chemist finds an hypothesis live enough to spend a year in its verification: he believes in it to that extent. But if his experiments prove inconclusive either way, he is quit for his loss of time, no vital harm being done” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 16). It is possible to read this as James just making a concession for the sake of argument; in a later passage, he claims that “science would be far less advanced than she is if the passionate desires of individuals to get their own faiths confirmed had been kept out of the game” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 27). Regardless, Marshall did not fit the description of James’ chemist. His research funding was nearing its end. Absent a breakthrough, he would have to pursue a job in private practice. Alternately, Marshall later wrote, “a successful infection with Helicobacter would point towards a career in clinical research, more exciting but likely to be financially insecure” (Reference Marshall and Grandin2006: 269).Footnote ¹¹ Marshall's option was, in James’ terms, both momentous and forced.

One might object that Marshall did not actually believe that H. pylori causes PUD. Rather, he merely found it sufficiently promising to risk experimenting on himself. Even if he and his contemporaries agreed that the hypothesis was possible but unlikely, further research into it was a bet with different payouts for Marshall than for his contemporaries. He was at a critical juncture in his career, but they were not. He might rationally take actions that they did not, like experimenting on himself, even without disagreeing about any beliefs. Thus, one might argue, they all had or ought to have had the same degree of belief in the hypothesis.Footnote ¹²

This objection can be tested by considering how Marshall and his colleagues acted in other contexts. The choice of how to treat a patient offered Marshall and his colleagues the same payouts, because the outcomes were just a third-party's condition. If he and his colleagues had the same degrees of belief, then they should have made the same treatment recommendations. Contrariwise, if Marshall believed his speculation that H. pylori causes PUD, then he should have been willing to act on it when the health of others was at stake. As it happens, Marshall was willing to make treatment recommendations even before his self-experiment. His recommendations were met with what Marshall describes as “a certain coolness” by his “more senior colleagues” (Reference Marshall and Grandin2006: 267). He was not allowed to treat ulcer patients with antibiotics before his self-experiment and subsequent publication. What stopped him from acting for non-research purposes was not a low credence in his hypothesis but instead social constraints of the clinical context. Those social constraints are important, because medical care should reflect standards of care rather than any doctor's idiosyncratic beliefs.

Ongoing research requires some scientists sometimes believing in advance of sufficient evidence. So scientific investigation in general and the discovery that H. pylori causes PUD in particular are counterexamples to the staunch evidentialists’ insistence that we ought never believe in advance of sufficient evidence.

Note that this is not a case of doxastic efficacy. Marshall's belief that H. pylori causes PUD made no difference to whether it does or not. I suppose we might isolate Marshall's belief ‘Future research will decisively show that H. pylori causes PUD’. This was doxastically efficacious: Because he believed it, the research got done. If he had not believed it, then the research probably would not have been done. Yet a belief about future research, like the belief about making friends in the previous section, readily separates into two components. Future research would decisively show that H. pylori causes PUD both because H. pylori does cause PUD and because the research would be conducted. Marshall's beliefs were efficacious in producing only the latter. The matter of real interest is that H. pylori causes PUD, and Marshall's beliefs had no influence on that one way or another.

So Marshall's case in particular and scientific investigation in general show that there are cases which lack doxastic efficacy but in which we should nevertheless “respect one another's mental freedom” (James Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 34). At a certain time, the evidence does not decisively speak for or against an hypothesis. There are risks attendant with all the options: believe it, reject it, suspend judgment and wait for further research. In such a situation, the costs and benefits of various possible outcomes guide scientists’ decisions about which option to adopt.

Following Carl Hempel (Reference Hempel1965), this is standardly called the argument from inductive risk. Since the argument applies to induction in the broad sense of ampliative inference, I will call it the argument from ampliative risk.Footnote ¹³ Reintroduced to recent philosophy of science by Heather Douglas (Reference Douglas2000, Reference Douglas2009), it is now taken as a standard argument for the claim that values may legitimately play a role in theory choice.Footnote ¹⁴

Although he did not apply it to scientific cases, the argument from ampliative risk is already given by James. The pursuit of knowledge demands both that we try to believe true things and that we try to avoid believing false things – as James writes: “Believe truth! Shun error!” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 25).Footnote ¹⁵ As evidence accumulates, the point at which one ought to believe is guided not by some impersonal function but by the appropriate balance between these two demands. James writes, “You … may think that the risk of being in error is a very small matter when compared with the blessings of real knowledge, and be ready to be duped many times in your investigation rather than postpone indefinitely the chance of guessing true” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 25–26).Footnote ¹⁶

In their characterization of the will-to-believe doctrine, A&T require both that the matter be “rationally undecidable” and that belief be “doxastically efficacious” (Reference Aikin and Talisse2018: 63, cited above). Thus, they miss the fact that ampliative risk and the efficacy of belief are independent arguments against staunch evidentialism.

Considerations of ampliative risk are in play not just in cases of mortal consequence but, as Gail Kennedy notes, “wherever one must act upon insufficient evidence” (Reference Kennedy1958: 584).Footnote ¹⁷ Since belief is always connected at least to possible action, though, ampliative risk is relevant wherever the evidence is less than compelling. For any live scientific question, the evidence admits of different interpretations and reasonable scientists may disagree. Ampliative inference necessarily goes beyond the evidence. There is always some ampliative risk, so there is always the potential for disagreement about how to weigh that risk – disagreement which may underwrite different judgments about what to believe.

It is not legitimate, though, to consider just any risk or consequence. The categorical goodness or badness of a state of affairs is irrelevant. In Marshall's case, the consideration in favor of believing that H. pylori causes PUD was not that it would be good for it to do so. Rather, the relevant considerations were what would to be gained by believing the hypothesis if it were true, what would be lost by believing if it were false, what would be lost by disbelieving if it were true, and what would be gained by disbelieving if it were false. Douglas (Reference Douglas2009) calls this values playing an indirect role. As she argues, theory choice always involves values in an indirect role but ought never involve values in a direct role. To put the point in different terms: Theory choice always involves the conditional values of believing/disbelieving given the hypothesis were true/false, but it should not involve the categorical value of the hypothesis being true/false.

In contrast, it is perfectly legitimate for values to play a direct role when belief is efficacious. In the alpine climber case where you are the climber, the value of your survival makes it reasonable for you to believe that you will survive.

So distinguishing the two considerations is not mere logic-chopping. Instead, there is an important sense in which ampliative risk is broader than doxastic efficacy, and another sense in which it is more narrow. Considerations of ampliative risk apply in a great many more cases, but only entitle us to assess beliefs in terms of specific conditional values (i.e., values may play only an indirect role). Considerations of doxastic efficacy apply to fewer cases, but entitle us to assess beliefs in terms of any personal factors (i.e., values may play a direct role).

To sum up the paper so far: James offers two kinds of considerations, and we can use these to distinguish two kinds of cases. Either would be sufficient for James’ positive argument that we may sometimes permissibly believe in the absence of compelling evidence.

One might think to defend Clifford by distinguishing epistemic reasons to believe from non-epistemic ones. The objection would go like this: Clifford's claim that one ought not believe when there is insufficient evidence can be construed as being just about epistemic reasons, and the considerations which James raises are extra-epistemic. Even if Jamesian concerns underwrite an all-things-considered conclusion that it is permissible to believe in the absence of sufficient evidence, this would be no counter-example to Clifford's thesis qua epistemological.

This objection fails for at least two reasons. First, neither Clifford nor James would accept such a distinction. The arguments that Clifford (and A&T) advance are not just concerned with getting the truth. As we'll see in the next section, they appeal to pragmatic consequences just as much as James does. Second, the considerations that James appeals to are epistemic. Regarding doxastic efficacy: The permission to believe comes precisely because Pr(S) is coupled to Cr(S) in such cases. Belief points to the truth. Regarding ampliative risk: Believing truth and shunning error are epistemic duties. Balancing between epistemic duties is still essentially an epistemic matter (Magnus Reference Magnus2018). Insofar as the balancing inescapably depends on your personal temperament, the epistemic inescapably includes such personal factors.

4. Hygiene and dogmatism

A&T – and Clifford before them – offer arguments that permissiveness about belief would be harmful. These worries are independent of whether James’ positive arguments for permissiveness are successful, so they merit at least some attention. I will summarize two arguments briefly and then rebut them.

The first argument is that believing permissively in some cases would weaken our habit of applying strict standards. In cases where it does matter – cases where James would agree that we should remain agnostic – weakened habits would leave us slipping up and believing where we should not. Call this the cognitive hygiene argument.Footnote ¹⁸

The second is that James’ argument is “a potent recipe for dogmatism” (Aikin and Talisse Reference Aikin and Talisse2018: 73). James concludes The Will to Believe by claiming that we ought “to respect one another's mental freedom” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 34). A&T object that this does not obviously follow. Instead, they note James’ focus on options which are momentous and irreversible. They formulate James’ argument as involving the commitment that “[i]f one believes fully for the sake of doxastic efficacy, one does not harbor doubts and one recognizes doubts as morally repugnant” (Reference Aikin and Talisse2018: 73). Rather than leading to respect for one another's mental freedom, A&T conclude, the upshot of James’ argument is that one should adopt a belief and cleave to it come what may. Call this the deepening dogmatism argument.

With respect to cognitive hygiene: A&T draw an analogy with dieting, suggesting that someone who is willing to believe on less than compelling evidence even once is apt to fall off the wagon. They write,

They are right, of course, that there is sometimes a slippery slope between doing something ‘just once’ and doing it many times. On some occasions I have suffered an upset stomach from eating a whole package of Twizzlers which I opened with the intention of eating just two. If the upset stomach were a rift in the very fabric of my life, then it would be better to never open Twizzlers – better still to wipe Twizzlers out of existence lest I eat too many again! Yet this is obviously absurd. I will eat the occasional Twizzler, and we can respect one another's gustatory freedom.

Moreover, the argument from cognitive hygiene fails to fit with important features of actual belief.

First, evidence is never unambiguously sufficient. Clifford supposes that induction is possible on the assumption that nature is uniform (Clifford Reference Clifford, Stephen and Pollock1886: 360). James notes that this assumption is not itself something which can be established with sufficient evidence.Footnote ¹⁹ We now recognize that the situation is even worse than that. After Goodman and the new riddle of induction, we know that an assumption of the uniformity of nature is insufficient to underwrite scientific inference. The real puzzle is not whether the future will resemble the past but how. Empirical inference is always ampliative. This means that it always involves risks. As we saw in the previous section, consideration of those risks makes it permissible for one to believe in some cases where others would find the evidence insufficient.

Second, the epistemic community is not as vulnerable as the objection presumes. Clifford argues not just that believing too quickly will lead down a slippery slope to credulity, but also that it “foster[s] a credulous character in others” (Reference Clifford, Stephen and Pollock1886: 345). As the example of Marshall and H. pylori illustrates, though, an epistemic community typically has mechanisms that allow for disagreement. Marshall believed an hypothesis that his fellows did not. Before more evidence accumulated, they tolerated his belief while giving it a cool response and refusing to let it guide treatment. Their tolerance did not mean that they were growing credulous. That specific case, as well as general considerations from philosophy of science, show that the community is structured so as to accommodate disagreement at least on questions which are being actively researched. Different epistemic communities handle dissent in different ways, but there is often at least a window in which individuals may productively disagree.

With respect to deepening dogmatism: A&T pose the objection as their own, but it resonates with concerns raised by Clifford who writes, “The danger to society is not merely that it should believe wrong things, though that is great enough; but that it should become credulous, and lose the habit of testing things and inquiring into them; for then it must sink back into savagery” (1886: 345).

In this form, the argument is no objection to James. Although James holds that we may permissibly believe in the absence of compelling evidence, this is in part because he is a fallibilist. He writes that “if we are empiricists, if we believe that no bell in us tolls to let us know for certain whether truth is in our grasp, then it seems a piece of idle fantasticality to preach so solemnly our duty of waiting for the bell” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 33–34). The fact that James calls this empiricism underscores the connection to further evidence. We ought always to go on testing things and inquiring into them. Belief should not stand in the way of further enquiry. James’ claim of a right to believe is the converse of this – there being further enquiry should not stand in the way of belief, either.

Of course, as a psychological matter, some evidence may be taken as compelling by a person on an occasion. As James writes, “The greatest empiricists among us are only empiricists on reflection: when left to their instincts, they dogmatize like infallible popes” (Reference James, Burckhardt, Bowers, Skrupselis and Burckhardt1979 [1897]: 22). In the instant before jumping, the alpine climber believes without hesitation. Yet the climber does not, as A&T write, treat “doubts as morally repugnant” (cited above). Taking a moment to disdain non-believers might spell doom as surely as doubting. And after the climber has survived, it will no longer be a matter of doxastic efficacy.