Part one: buck-passing and reductionism

Goldberg characterizes deferral of challenges, or buck-passing, as pointing to a further source of justificatory reasons for a subject’s belief. If challenged to justify a testimonial belief, the subject will make the case for accepting the testimony of a given speaker. But when her own justificatory resources are exhausted, she can tell the challenger that if he wants more in the way of justification, he can simply ask the original speaker. This is what Goldberg calls ‘passing the buck.’ In so doing, he says, one ‘points in the direction where further epistemic support can be found (135).’

What makes it appropriate to pass the buck, Goldberg says, is that reasons for thinking the speaker trustworthy on the present occasion are ipso facto reasons for thinking that further support for the claim at issue can be found. He notes that buck-passing is possible when we rely on human beings, as opposed to gages or instruments, since a rational being shapes its beliefs to fit the evidence it has. While we may rely on mere instruments, we cannot appeal to their reasons.

Goldberg makes two claims about buck-passing, so understood. The first is that it is only in cases of testimony that it is epistemically appropriate to pass the buck. The second is that reductionists about testimony can agree that there is a right to pass the buck. He concludes that even a reductionist can allow that testimonial knowledge is a distinct kind of knowledge. For the reductionist can say testimonial knowledge is distinct, not because there is a distinctively testimonial kind of justification, but because testimony is the only source of knowledge in which one has the right to pass the buck.

By contrast, we do not have the right to defer challenges when our source is a mere instrument, like a thermometer. This is because thermometers are not ‘rational beings.’ As Goldberg puts it, ‘rational beings rely on … evidence of the sort that reliable instruments merely offer (136).’ And this difference, he says, explains why one has the right to pass the buck to a speaker, but not to a gage or instrument.Footnote ¹

Suppose for instance that I see a thermometer that reads 44°F and on that basis form the belief that it is 44°F. If I then tell you it is 44°F, and you don’t immediately accept what I say, I can tell you about the thermometer, and I can give my reasons for thinking the thermometer’s reading is a reliable indication of the temperature. Once I have done that, I have said all I can say to defend my claim. But in the case where someone, call her S, has told me the temperature, a further move is available. As with the thermometer, I can tell you about S’s assertion. And I can give my reasons for thinking S’s assertion is a reliable indication of the temperature. But that is not all I can do. In addition, I can pass the buck to S. I can say, ‘If you want more in the way of a defense of the claim that it is 44°F, you can simply go ask S.’

For it to be epistemically appropriate to pass the buck, Goldberg holds, one must be justified in believing that if the challenger wants more evidence, he should consult the original speaker. This comes out in his discussion of another example:

Seeing that that Dien wants more in the way of a defense of the claim that the Yankees won, Anita replies with a ‘move of last resort.’ She says,

Goldberg says this move of last resort is epistemically appropriate. In support of this he notes that Anita is justified in her belief that Brad will have additional reasons:

Moreover, it is because Anita is justified in believing Brad’s assertion reliably indicates the truth that she is justified in believing it is to Brad that Dien ought to be looking, if he wants more evidence. Her reasons for believing the former, recall, are ipso facto reasons for believing the latter. So what makes it epistemically appropriate for Anita to send Dien to Brad is that she has reasons of her own that justify her in believing the Yankees won last night. When Dien asks her how she knows this, she can defend her claim to knowledge just by saying Brad told her and giving her reasons for thinking him reliable. So she can send Dien to Brad for more evidence without thereby undermining her claim to know. As Goldberg puts it, the subject is epistemically entitled to make the move of last resort ‘only if she has already vindicated her epistemic right to rely on the testimony in question (Reference Goldberg, Jennifer and Ernest2006, 137).’

The sense in which Anita ‘can’ tell Dien that if he wants more evidence he should go ask Brad is that Anita can say this to Dien, and be expressing a justified belief, without thereby undermining her claim to know that the Yankees won last night. This is often possible when one’s source is a rational being. But according to Goldberg, it is not possible when one’s source is a mere instrument.Footnote ² To say, ‘If you want more evidence, you should consult the thermometer, ’ would be to assert something one was not justified in believing, since one is not justified in believing the thermometer is relying on evidence that one lacks.Footnote ³ Thermometers don’t rely on evidence; they only offer it.

So there is a sense in which one can pass the buck to one’s source, when one’s source is a rational being. If we let p be some proposition that A asserts, we can put this sense as follows:

When A’s source is a rational being, Weak B-P may be true. We will later contrast Weak B-P with a stronger thesis we call Strong B-P.

Now we ask whether Weak B-P is acceptable to the reductionist. Rather than trying to say what reductionism is, we take over Goldberg’s characterization of it. We ask whether the reductionist, so characterized, can accept that there is a weak right to pass the buck when one’s source is a rational being.

As Goldberg sees it, reductionists and anti-reductionists agree that to have testimonial knowledge, one must have an epistemic right to rely on the speaker’s testimony. On neither view is it enough that the speaker’s testimony actually be reliable (Reference Goldberg, Jennifer and Ernest2006, 128–129). The reductionist and the anti-reductionist part company, though, over what it takes to have an epistemic right to rely on the speaker’s testimony. The reductionist says one has this right only if one has ‘reasons (including those of a standard inductive sort)’ to think that the speaker’s assertion is a reliable indication of the truth:

The anti-reductionist, by contrast, says that the justification of A’s belief in what S says does appeal to a testimony-specific epistemic principle. For instance, the anti-reductionist might claim that there is a ‘defeasible but presumptive right to trust other speakers.’ This would be the sort of testimony-specific principle that the reductionist refuses to acknowledge.

Can one accept that Weak B-P is sometimes true, without acknowledging principles of justification that are specific to testimony? If so, then the reductionist can accept that Weak B-P is sometimes true.

Goldberg argues that the reductionist can indeed accept this:

In other words, the reductionist can say that it is only because Anita is already justified in relying on Brad’s testimony that she is also justified in believing Brad has evidence that she lacks. What justifies her in relying on Brad’s testimony, the reductionist can say, is that she has reasons to think his assertion is a reliable indication of the truth of p. And the reductionist is free to insist that these reasons, if they are to justify Anita in relying on Brad’s testimony, must rest on principles that apply quite generally. The reductionist could insist, for example, that these reasons be ordinary inductive ones. Provided that these reasons justify Anita’s belief that the Yankees won last night, Anita can go on to admit that Brad has evidence she lacks, without thereby undermining her claim to know that the Yankees won last night.

There is thus nothing in Weak B-P that requires an appeal to testimony-specific principles of justification. No such principles are required to explain how Anita can admit that Brad has evidence that she lacks, without thereby undermining her claim to know that the Yankees won last night. The reductionist can therefore accept that Weak B-P is sometimes true.

We agree that reductionists can acknowledge a weak right to pass the buck. But Goldberg relies on a further claim. He holds that this right arises only in cases of testimony (Reference Goldberg, Jennifer and Ernest2006, 137, 139). This is why he thinks the reductionist can accept that testimonial knowledge is knowledge of a distinctive kind.

We will now argue that the weak right to pass the buck is not unique to cases of testimony.

In the following example I have knowledge that p because I know that another individual, N, knows that p. If challenged, I have reason to point to or refer to N as a further source of justification. But this is not a case of testimony. In the example, no one has told me that p.

This example shows that weak buck-passing is not unique to testimonial knowledge. In the example, I have cited my source and given my reasons for thinking the source is reliable. And I can tell my interlocutor, and thereby express a justified belief, that if he wants more evidence he should go ask the source. Moreover, I can do this without undermining my claim to know that the enemy will be at L at t. Weak B-P is thus true in this case. But my source has not told me where the enemy will be at t. Indeed we can suppose my source has not told anyone where the enemy will be at t. My knowledge, therefore, is not testimonial. If there is something distinctive about testimonial knowledge, Weak B-P does not capture it.Footnote ⁴

Goldberg anticipates the worry that buck-passing, as he conceives it, is not unique to testimony. He considers the following case:

McSorley’s knowledge that there was a red car in Richardson’s driveway is not testimonial. It is grounded in perception and memory. And yet McSorley can send Richardson to the experts for more evidence. ‘If this is a case of epistemic buck-passing, ’ Goldberg says, ‘it challenges the idea that the phenomenon of epistemic buck-passing is what singles testimonial knowledge out as a distinctive epistemic kind (Reference Goldberg, Jennifer and Ernest2006, 139).’

Goldberg replies that the evidence to which McSorley points, in sending Richardson to the experts, does not directly support the claim that there was a red car in Richardson’s driveway. It directly supports a different claim, namely that perception and memory are reliable belief-forming processes. But it supports the claim about the red car only indirectly, by way of the claim about the reliability of perception and memory. This prompts Goldberg to qualify his original claim. What is unique to testimonial knowledge, he says, is that one can send one’s interlocutor to one’s source for evidence that directly supports one’s testimonial belief:

But BRITISH INTELLIGENCE shows that even this is false. In that example, the evidence to which I point is a direct source of support for the content of my belief. I do not point to evidence that the intelligence agency or its methods are reliable, as McSorley points to evidence that memory and perception are reliable. Rather, I point to evidence that the enemy will be at L at t, evidence that I take the agency to have acquired by breaking the enemy code. Even with Goldberg’s qualification, the weak right to pass the buck is not unique to testimonial knowledge. Goldberg’s argument does not show that the reductionist can acknowledge testimonial knowledge as a distinctive epistemic kind.

We’ve seen that the reductionist can embrace weak buck-passing. But there is a stronger claim that the reductionist cannot accept. The stronger claim is that even if she does not have enough evidence to justify her belief, Anita may know the Yankees won last night. Suppose Anita is unable to justify her belief by herself, and is forced to turn to Brad for help. In that case, Anita’s belief is justified in part by evidence that only Brad has. If that were so, then there would be a testimony-specific principle of justification. The principle would say that testimonial belief can be justified in part by reasons one does not possess, but that the speaker does possess.

On this stronger reading, to say I have a right to pass the buck is to say that if someone challenges me to justify my belief, I can indicate that I cannot do so without help from the speaker, without thereby undermining my claim to knowledge.Footnote ⁵ If we let p be some proposition that A asserts, then we can put the strong sense like this:

This is stronger than the weak reading, because the weak reading does not say A can acknowledge that she does not possess enough evidence to justify her belief. It says only that A can tell the hearer to go ask the source. For me to say, ‘If you want more evidence you should go ask my source, ’ is not to admit that I can’t justify my belief that p all by myself. For it may be that by giving what justification I can, I have justified my belief all by myself. It may be, to go back to the example, that by telling Dien that she heard it from Brad, and giving her reasons for thinking Brad’s assertion is reliable, Anita has in fact justified her belief that the Yankees won last night all by herself. In that case Dien does not need more evidence, in order to see that Anita is justified. He may want more evidence. Perhaps he grants that she is justified, but still wonders if her justified belief is true. In that case, Anita can tell him Brad will have more evidence. In so doing, she is not admitting that she can’t justify her belief all by herself.

Indeed it is this feature of Weak B-P that Goldberg exploits, to show that the reductionist can acknowledge the epistemic right to pass the buck:

Goldberg’s point here is that the reductionist can say Anita has justified her belief that the Yankees won last night just by citing the reasons she has. The reductionist can thus regard Dien’s request for more evidence not as a challenge to Anita’s epistemic right to rely on Brad’s assertion, but as granting that right and asking the further question whether her justified belief is true. And the reductionist can allow that it is perfectly appropriate for her to answer that further question by sending Dien to Brad.

But suppose Dien wants to see the evidence that justifies Anita’s belief that the Yankees won last night. In that case Dien is not granting that Anita’s belief is justified and going on to ask a further question. Rather, he is challenging her to justify her belief. If Anita defers to Brad on that question, she is admitting that the evidence needed to justify her belief is not all in her possession. And if she can admit that without thereby undermining her claim to know that the Yankees won last night, then Strong B-P is true in her case.

The question we wanted to discuss was whether a reductionist could accept that testimonial knowledge is a distinct kind of knowledge. Goldberg argues that the right to pass the buck is unique to testimonial knowledge, and that a reductionist can accept that there is a right to pass the buck. The availability of the ‘move of last resort’ is thus a distinguishing feature of testimonial knowledge, even if reductionism is true. We contend, however, that the argument fails on both the weak and the strong readings of buck-passing. The weak right to pass the buck is acceptable to the reductionist, but is not unique to testimonial knowledge. One can have this right even when one has not been told that p, as we saw from BRITISH INTELLIGENCE. Conversely, the strong right is plausibly unique to testimonial knowledge, but is not acceptable to the reductionist.

Recall that Weak B-P is true in BRITISH INTELLIGENCE. Noticing that the British have repeatedly succeeded in stationing forces just where the enemy is going to be, I infer that they have cracked the enemy’s code. Upon learning that the British forces will be at L at t, I conclude that the enemy will be at L at t. Here, Weak B-P is true. When I give you my evidence for thinking the enemy will be at L at t, you might accept that I am justified in believing that, but go on to wonder whether my belief, justified as it is, is true. In that case I can say that if you want more evidence you should go ask the British, without thereby undermining my claim to know that the enemy will be at L at t.

But Strong B-P is not true in this case. If you ask to see the evidence that justifies my belief, I can’t admit that I am unable to produce such evidence, and send you to the British. If I do that, I undermine my claim to know where the enemy will be. Instead, I need to produce the evidence that justifies me in believing that British troop movements reliably indicate where the enemy will be. Since no one has told me where the enemy will be, I have no choice but to go on the evidence I have.

This example shows that weak buck-passing is not unique to testimonial knowledge. But it does not touch the claim that strong buck-passing is unique to testimonial knowledge.

Moreover it is plausible that Strong B-P is unique to testimonial knowledge. For according to Strong B-P, the evidence the subject has may not suffice to justify the subject’s belief, the difference being made up by evidence possessed only by the source. Outside of a testimonial exchange, however, it is hard to see how the subject could escape the burden of having enough evidence to justify her belief. One might put this thought in terms of epistemic responsibilities. Testimony may well be an act in which one assumes partial responsibility for the justification of the hearer’s belief (Moran Reference Moran2005; McMyler Reference McMyler2011, 68–76; Owens Reference Owens2006). But where there is no testimony, it is hard to see how the subject could avoid being wholly responsible for having the evidence that justifies her belief.

Unlike Weak B-P, Strong B-P is plausibly unique to testimonial knowledge. So if the reductionist could accept that Strong B-P is sometimes true, Goldberg could use Strong B-P to argue that the reductionist can accept that testimonial knowledge is a distinct kind of knowledge.

But in fact the reductionist cannot accept that Strong B-P is sometimes true. Reductionism, recall, is the view that there are no testimony-specific principles of epistemic justification. If Strong B-P is sometimes true, however, then there is a principle to the effect that testimonial belief can be justified even when the believer does not have all the evidence that justifies her belief. And if strong buck-passing is unique to testimonial knowledge, then this is a testimony-specific principle. It applies specifically to cases where one has been told. In other cases, such as BRITISH INTELLIGENCE, one’s belief is not justified unless the evidence that justifies it is in one’s own possession.

We think anti-reductionists can use Strong B-P to argue that reductionism fails to capture the distinctiveness of testimonial knowledge. But any such argument will rely on the claim that Strong B-P is at least sometimes true. How might an anti-reductionist support this claim?

One strategy would be to recast existing anti-reductionist arguments as arguments for Strong B-P.

For example, anti-reductionists often appeal to the case of testimonial knowledge in children. Before a certain age, they say, children cannot assess speakers for trustworthiness.Footnote ⁷ But often, when they accept what they are told, they wind up with knowledge. ‘Information and knowledge simply seem to flow in, ’ as Burge puts it (Reference Burge2013, 265). One might argue that children cannot shoulder the responsibility of assessing grounds for belief, and that it is enough, for their knowing, if someone else has done the assessing for them.Footnote ⁸ Thus children can have testimonial knowledge even if they lack the evidence required to justify their beliefs on their own, as long as their adult sources have their backs, so to speak. And that is the idea behind Strong B-P.

There is a difficulty with this argument as it stands. One could agree that children need not be able to assess the grounds for their beliefs, and still insist that sufficient grounds must be available to them. Perhaps children can only have testimonial knowledge when they are told things that are in fact justified on the basis of evidence that is somehow in their possession. In that case, the strong buck-passing thesis is still false. But it is hard to see why this restriction should hold. Suppose a child can know the time because I tell him, despite his being unable to read clocks, and despite his being unable to assess the trustworthiness of my testimony. Why should it matter whether the clock I go on is in his visual field?

For another example, consider McMyler’s argument for the buck-passing thesis. McMyler argues that a hearer has an epistemic right to defer challenges back to the original speaker. His argument is that when a hearer is challenged to defend a claim she has accepted on testimony, and cannot do so, she need not immediately ‘give up her claim to know.’ Instead she can defer the challenge back to the original speaker (McMyler Reference McMyler2011, 75). If the original speaker cannot meet the challenge, Footnote ⁹ then the hearer does have to give up her claim to know. The point is that the question of knowledge is not settled by the fact that the hearer cannot defend her belief on her own. If, in order to have knowledge, the hearer must be able to defend her testimonial belief on her own, without help from the speaker, then the question of knowledge is settled by the fact that the hearer cannot defend the claim on her own. In fact, however, we treat the hearer’s inability to meet the challenge as inconclusive. We do not immediately reject her claim to know. Instead, we wait to see whether the speaker can meet the challenge. On its face, this too supports Strong B-P. For Strong B-P says the hearer need not give up her claim to know just because she cannot justify her testimonial belief on her own. Instead she can send the challenger to the source.

Our conclusion is not that these arguments are decisive, but that they make for a sympathetic rendering of the case against reductionism. The charge against reductionism is that it fails to capture what is distinctive about testimonial knowledge. For a charitable reading of this charge, we need to look for something that is arguably true of testimonial knowledge, that is true only of testimonial knowledge, and that is unacceptable to the reductionist. Strong buck-passing fits the bill on all three counts.

But while strong buck-passing is stronger than weak buck-passing, it is still weaker than much of what anti-reductionists have said about testimonial knowledge. This means there is an additional advantage to the anti-reductionist strategy we have in mind; it avoids commitments that expose some forms of anti-reductionism to sharp objections. For example, some anti-reductionists are committed to the idea that features of the relationship between speaker and hearer generate a special non-evidential sort of epistemic reason. As we shall see in the next section, this exposes them to a serious dilemma. But Strong B-P carries no such commitment.

Part two: strong B-P and the interpersonal view of testimony

The question of reductionism and anti-reductionism is the central issue in the epistemology of testimony. In recent years some anti-reductionists have emphasized the interpersonal, or second-personal, nature of testimony (Hinchman Reference Hinchman2005; Moran Reference Moran2005; McMyler Reference McMyler2011). They have depicted testimonial knowledge as arising from a kind of transaction in which the speaker offers an assurance, or extends an invitation to trust, which the hearer then accepts. And they have insisted that reductionists cannot do justice to the social nature of the knowledge that results from this sort of exchange.

To press this point they must articulate the idea that there is something distinctively social about testimonial justification. The standard way of doing this is to say there is a special kind of epistemic reason that arises from the relationship between speaker and hearer. As Jennifer Lackey has pointed out, however, this makes the interpersonal view look distinctly unpromising. For then the view faces a dilemma. Either it says interpersonal factors confer justification on testimonial belief regardless of whether the testimony is reliable, or it adds reliability as a condition on testimonial justification. On the first horn, the interpersonal factors are powerless to confer justification on the belief. On the second, the interpersonal factors are superfluous, all the epistemic work being done by the fact that the testimony is reliable.

We believe that the strong notion of buck-passing articulated above provides an alternative to the standard way of articulating the social character of testimonial justification. Strong B-P captures the idea that the justification of testimonial belief is a social rather than an individual achievement. But Strong B-P does not imply that there is any special kind of epistemic reason or evidence that arises from a testimonial exchange. This way of understanding the interpersonal view of testimony does not fall prey to Lackey’s dilemma. To show this, we begin with a look at the dilemma.

In Chapter 8 of Learning from Words Lackey raises an objection to what she calls the Interpersonal View of Testimony, or IVT. Proponents of this view, she says, agree on the following three claims:

• (1) The interpersonal relationship between the two parties in a testimonial exchange should be the central focus of the epistemology of testimony.

• (2) Certain features of this interpersonal relationship – such as the speaker offering her assurance to the hearer that her testimony is true, or the speaker inviting the hearer to trust her – are (at least sometimes) actually responsible for conferring epistemic value on the testimonial beliefs acquired.

• (3) The epistemic justification or warrant provided by these features of a testimonial exchange is non-evidential in nature (Reference Lackey2008, 221).

For a practice to have epistemic value, Lackey explains, is for it to ‘somehow further the goal of acquiring true beliefs and avoiding false ones (Reference Lackey2008, 226).’ Thus the IVT is committed to the idea that certain non-evidential features of the relationship between the two parties somehow further the goal of believing the truth. Proponents of the IVT must therefore explain how such relations as telling and being believed, or assuring and accepting, can confer epistemic value on the belief that p.

According to Lackey, the IVT faces a dilemma at this point. There are two ways to explain how these features confer epistemic value on the resulting belief. One way is to say these features further the goal of truth regardless of whether the speaker is reliable, and regardless of whether the hearer is rational in accepting the testimony. The other is to say that these features further the goal of truth as long as further conditions, such as reliability and rationality, are also satisfied. On the first horn, the IVT is too permissive. In cases where the speaker is unreliable or the hearer is epistemically irresponsible, such features as assurance and acceptance do not further the goal of truth. Consequently the IVT ends up acknowledging justification or warrant where, from an epistemic point of view, it should not. On the other horn, we no longer have a genuinely interpersonal theory. This is because the additional conditions of reliability and rationality now do all the work of conferring epistemic value on the hearer’s testimonial belief.

To illustrate the first horn, Lackey gives us the case of Vera. Vera is ‘a radically unreliable believer who consistently offers assertions to her hearers that she sincerely believes to be true but which are wholly disconnected from the truth (Reference Lackey2008, 227).’ Lackey points out that while Vera may assure us, invite us to trust her, offer to take responsibility for the truth of what she says, or guarantee that what she says is true, none of this will further the goal of true belief. This is because Vera’s assurances, invitations, guarantees and offers are disconnected from the truth. She is as likely to assure us of something false as of something true, so we make no advance on the truth by accepting her assurances.

To fix this problem, one can add conditions to the IVT. Following Hinchman, for example, one might say that in order for the acceptance of an invitation of trust to confer epistemic value on one’s testimonial belief, the speaker’s testimony must be a reliable guide to the truth, and the hearer must have no evidence that the speaker is untrustworthy.Footnote ¹⁰ But while this adjustment puts the interpersonal view ‘on the epistemological map, ’ Lackey says, it does so by making trust superfluous.

To illustrate this second horn, Lackey gives us the case of Ben, Kate and the eavesdropper Earl. Ben tells Kate that their boss is having an affair with another co-worker, Irene. Unbeknownst to them, Earl overhears their conversation and now believes solely on the basis of Ben’s testimony that his boss is having an affair with Irene. We can assume that the boss is having an affair with Irene, that Ben’s testimony is a reliable guide to this truth, and that Kate has no evidence that Ben is untrustworthy. In that case, Hinchman’s additional conditions are satisfied for Kate, and Kate’s acceptance of Ben’s invitation to trust should confer epistemic value on her belief that her boss is having an affair with Irene.

Lackey argues, however, that according to any version of the IVT, Kate’s testimonial belief must possess epistemic value that Earl’s does not. This is because Ben has not invited Earl to trust him, or offered his assurance to Earl. Earl is therefore not in a position to accept Ben’s invitation to trust. So if Kate’s acceptance of Ben’s invitation to trust confers epistemic value on Kate’s belief, her belief must have epistemic value that Earl’s lacks.

But what, Lackey asks, could distinguish Kate’s and Earl’s beliefs epistemically? We can assume that Earl, like Kate, has no evidence that Ben is untrustworthy. We can also assume that Kate and Earl are equally well acquainted with Ben, that neither is more gullible than the other, and that they possess the same background information about the proposition that Ben asserts. It is true that Kate’s belief arises from her acceptance of an invitation to trust, whereas Earl’s does not. But how does that put Kate in a better position to believe the truth?

One answer is that if Ben’s claim turns out to be false Kate has grounds to feel let down, or to complain or feel resentment, whereas Earl does not. Another is that Ben can rightly feel slighted if Kate does not believe him, whereas he cannot rightly feel slighted if Earl does not believe him. But Lackey responds that even if these claims are true, these entitlements to various reactions ‘do not bear in any way on the truth-conduciveness or epistemic rationality of the testimonial beliefs in question’:

On the second horn, features like assuring and accepting, or inviting to trust and trusting, in fact do nothing to further the goal of believing the truth. All of the epistemic work is done by the additional conditions concerning reliability and the absence of contrary evidence. ‘In providing the epistemic explanation of B’s newly acquired justified or warranted belief, then, trust in A simply drops out of the picture (Reference Lackey2008, 238)’.Footnote ¹¹

One way of trying to articulate the social character of testimonial justification is to say that interpersonal factors somehow add epistemic value to the hearer’s belief. One might say, for instance, that testimonial contexts give rise to non-evidential reasons (Moran Reference Moran2005), or second-personal evidence (McMyler Reference McMyler2011), Footnote ¹² supporting the truth of the hearer’s belief. We believe Lackey has raised a serious difficulty for these views. But Strong B-P articulates a different way in which testimonial justification might be social. And this means the IVT as Lackey defines it does not exhaust the field of interpersonal views. Not every interpersonal view of testimony says there are features of the testimonial relation that add epistemic value to the hearer’s belief, by providing epistemic justification that is non-evidential in nature. Consider the strong reading once more:

Notice that Strong B-P says nothing about the kind of evidence, or epistemic reason, that justifies A’s belief that p. It says only that A’s claim to know that p is not impugned by the fact that she cannot produce all of the evidence that needs to be produced in order to show how her belief is justified. It is therefore open to hold that all of the reasons that justify the hearer’s belief are of an ordinary evidential sort. What Strong B-P says is just that A need not possess all those reasons in order for her belief to count as knowledge.Footnote ¹³ It does not say there is a non-evidential kind of epistemic reason, nor does it say there is a special second-personal kind of evidence.

On the other hand, as we have seen, Strong B-P allows us to state the sociality condition in a way that is unacceptable to the reductionist. To say of a belief that its justification is a social achievement, on this construal, is to say the evidence that justifies the belief is not all in the subject’s possession, some of it being in the possession of another person. If we add that this can happen in testimonial contexts, and only in testimonial contexts, then it follows that there is a testimony-specific principle of epistemic justification. The principle says testimonial belief can be justified even when the believer does not have all the evidence that justifies her belief.

Read this way, the sociality condition does not entail that features of the interpersonal relationship between speaker and hearer confer epistemic value on the hearer’s belief. It does not say that an assurance or an invitation to trust provides a reason in favor of the truth of the hearer’s belief, or adds to its justification or warrant. Rather, it says that when a hearer believes a speaker, this can affect the hearer’s epistemic responsibilities. Specifically, it can relieve her of the responsibility of possessing all the evidence that justifies her belief. This is not to say that the mere fact of believing someone, or taking someone’s word, does anything to justify the belief. It is merely to say that the belief may be justified even if the hearer does not have all the evidence needed to justify it, and must therefore turn to the speaker for help.

While Lackey has a sharp objection to the IVT as she defines it, her dilemma does not affect this version of the interpersonal view.

To see this, recall how the dilemma arises. On one horn, the IVT acknowledges justification or warrant where, from an epistemic point of view, it should not. This is because, on this horn, the IVT says an assurance or invitation of trust confers warrant or justification on the hearer’s belief regardless of whether the speaker is reliable. The case of Vera shows that an assurance or invitation of trust alone does not add any justification or warrant to the hearer’s belief. Because Vera is consistently unreliable, her assurances (however sincere) add no warrant at all. So if the IVT says assurance adds warrant regardless of the reliability of the speaker, then it sees warrant where it should not.

On the other horn, we no longer have a genuinely interpersonal theory. This is because, on this horn, the IVT says an assurance or invitation of trust confers warrant or justification on the hearer’s belief only if the speaker is reliable. The case of Ben and Kate shows that when the speaker is reliable, no warrant is added by the assurance or invitation to trust. Ben has extended an invitation of trust to Kate, but not to Earl. If this invitation added justification or warrant to Kate’s belief, she would have some warrant that Earl lacks. But Earl has just the same warrant or justification as Kate.

As Lackey sees it, this will mean the theory is no longer genuinely interpersonal. A genuinely interpersonal theory, in Lackey’s view, must be one on which assurance or inviting trust adds warrant to the hearer’s belief. It is this assumption that we want to question.

As we have seen, there is a way of understanding the interpersonal view on which it does not say assurance adds warrant or justification to the hearer’s belief. On this version of the view, what the testimonial context does is not to add warrant but to alter the hearer’s epistemic responsibilities. When a hearer believes a speaker, this can exempt the hearer from the normal responsibility to possess the evidence that justifies one’s beliefs. The hearer can thus admit that she lacks this evidence, and turn to the speaker for help, without thereby undermining her claim to knowledge.

Notice also that this version of the interpersonal view says nothing about address. In Lackey’s example, Kate is addressed and Earl is not. But the strong buck-passing thesis is silent on the question whether the hearer must be addressed. So this version of the interpersonal view does not predict that Kate and Earl are in epistemically different positions. In developing a view of this kind, one is free to go either way. One might argue that Earl, unlike Kate, is on his own. That is, one might say it is up to Earl alone to possess the evidence that justifies his belief, and that otherwise his claim to knowledge is undermined. Alternatively, one might argue that Earl too can admit that he lacks the evidence that justifies his belief, and send challengers to Ben, without thereby undermining his claim to know.Footnote ¹⁴ In that case the features of testimony that affect the hearer’s epistemic responsibilities are not those of being told, assured, or invited. They are more general. For instance, one might say the mere fact that the speaker tells someone is enough to effect this shift in responsibilities for anyone who hears the testimony, including those who are not addressed. Or more radically, one might deny that there need be any addressee at all, as in the case of a diary or soliloquy. We take no position on this here. The strong buck-passing version of the interpersonal view is consistent with a wide range of answers to this question.

The neutrality of Strong B-P on the question of the epistemic significance of address has an important implication. It might seem that one could escape Lackey’s dilemma just by showing that Kate’s belief is warranted in a way that Earl’s is not. The idea would be that if address makes an epistemic difference, then it must do so by creating non-evidential epistemic reasons. But now we can see that this does not follow. One way to understand how address makes an epistemic difference is to say it is only addressees who acquire a strong right to pass the buck (McMyler Reference McMyler2013). But one can say this without positing non-evidential epistemic reasons. To show that address is epistemically significant, therefore, is not necessarily to vindicate the IVT as Lackey understands it.

Hinchman, for example, holds that testimony ‘makes a warrant available’ to the addressee that it does not make available to unaddressed overhearers (Reference Hinchman2014). In telling the addressee that p, the speaker assures the addressee that p has the epistemic support the addressee needs it to have. The testimony is thus tailored to the epistemic needs of the addressee, whose needs may differ from those of overhearers. This idea fits well with the strong buck-passing version of the interpersonal view. One might argue that the warrant thus ‘made available’ to the addressee consists in evidential reasons possessed by the speaker. And if Hinchman is right that this warrant is not made available to unaddressed overhearers, one might conclude that only addressees can have a strong right to pass the buck. Strong B-P would thus be true of Kate but not of Earl. Again, such a view is not interpersonal by Lackey’s lights, because it does not posit non-evidential epistemic reasons. It is interpersonal by our lights, however, because it holds that Strong B-P is sometimes true.

At this point one might wonder in what sense the strong buck-passing view is still interpersonal. The answer is that it still embodies the sociality condition. It still says that testimonial justification can be a social achievement, since the evidence that justifies the hearer’s belief can be distributed across the testimonial relationship, some of it being possessed only by the speaker. It thus rejects the assimilation of testimony to cases like BRITISH INTELLIGENCE, in which the hearer cannot admit that she lacks the evidence that justifies her belief, without impugning her claim to knowledge. In those cases, there is a weak right to pass the buck, since the evidence that justifies the hearer’s belief also justifies her in believing the speaker will have more evidence than she herself has. But there is no strong right to pass the buck, because the hearer must be in possession of the evidence that justifies her belief.

Interestingly, there is a real question whether Lackey’s own ‘dualist’ theory counts as interpersonal by this test. On Lackey’s view the hearer’s possession of positive reasons for believing the testimony is necessary, but not sufficient, for the hearer’s being justified in believing p. It is not sufficient because, in addition, the speaker must be reliable (Reference Lackey2008, 176–177). But the reliability of the speaker is not something the hearer can supply. So testimonial justification is a social achievement in the sense that the speaker satisfies the reliability condition and the hearer satisfies the evidence condition.

However, Lackey retains the idea that the hearer must be in possession of the evidence that justifies her belief. On her view, what is needed for justification is not more evidence but reliability. Thus the hearer must have enough evidence to justify the belief. Lackey’s dualism rules out the strong buck-passing thesis. It is not interpersonal by that standard.

Lackey could respond that in fact no amount of evidence possessed by the hearer is enough to justify the belief, because a further condition has to be satisfied, namely the reliability condition. So in fact, Lackey could say, her dualism implies that a testimonial belief can be justified even if the hearer does not have enough evidence to justify the belief. Indeed this appears to be the only way a testimonial belief can be justified, on her view.

This response founders, however, on an ambiguity in the notions of sufficiency and insufficiency. It is one thing to say the hearer has enough evidence to justify her belief. It is another to say that having that amount of evidence is a sufficient condition of her belief’s being justified. The former can be true even if something other than evidence is also required. For example, there is no amount of flour such that having that amount of flour is a sufficient condition of baking a cake. One will also need enough eggs, enough time and so on. But there is an amount such that no more flour is needed. It is possible, therefore, to have enough flour to bake a cake, without having everything one needs to bake a cake.

On Lackey’s view, there is no amount of evidence such that having that amount of evidence is a sufficient condition of having a justified testimonial belief. But it is still possible for a hearer to have enough evidence to justify her testimonial belief. It is possible, in other words, that no more evidence is needed, just as it is possible that no more flour is needed. So Lackey’s dualism does not imply that a testimonial belief can be justified even if the hearer does not have enough evidence to justify it. In fact, it implies that the hearer must have enough evidence to justify the belief. If she doesn’t, then more evidence is needed, and Lackey does not allow that a belief can be justified by evidence that the hearer does not have.

The strong buck-passing thesis, on the other hand, implies precisely that a belief can be justified by evidence that the hearer does not have. The hearer can admit that more evidence is needed, and turn to the speaker to make up the difference, without thereby undermining his claim to know. This shows that Lackey’s dualism rejects the strong thesis. Her view is not interpersonal by this measure.Footnote ¹⁵

Notes on contributors

Professor Judith Baker taught at Glendon College, York University at the time of her death in 2014. She was well known for her work on trust and the rationality of belief.

Philip Clark is Associate Professor at the University of Toronto. His interests include the practical syllogism, deliberation, belief, intention and practical knowledge.