1

Virtually all parties in the epistemology of testimony agree that testimony is a significant source of justified beliefs and knowledge.Footnote ¹ We cannot do without testimony; otherwise we would be limited in our access to the world to just those parts of it that fall before our senses, have been recorded in our memories, are subject to introspection, a priori intuition, or which can be inferred from the aforementioned sources of evidence. Yet we do know more than what we have simply from these sources. Testimony furnishes us with many or most of our justified beliefs and knowledge about science, history, geography, politics and the contents of the minds of others. I will follow many currently working in the epistemology of testimony in taking observations like the above to constitute a constraint on theorizing about the epistemology of testimony. If, according to a theory of testimony, it turns out that most of us know very little via testimony, so much the worse for the theory. If, in the actual world, a theory of testimonial evidence entails that we do not know much of what it seems to us that we do know via testimony, that theory is false. Call this the ‘non-skeptical constraint’ (Weiner Reference Weiner2003: 256).

Testimony furnishes us with evidence. Suppose I approach a stranger and ask where the nearest coffee shop is. If she tells me, ‘There's a café just around the corner on the left’, I will often thereby come to be justified in believing what I am told. If my friend, the Revolutionary War enthusiast, is giving me a tour of Valley Forge and says to me, ‘George Washington slept here’, then I may even be in a position to know that George Washington slept here on the strength of hearing him say this. At least, we may be justified or in a position to know what we are told if we do not possess defeaters which undermine or counterbalance what we receive through testimony. Call ‘testimonial evidence’ the good epistemic reasons for believing that p we receive in virtue of awareness or justification for believing that someone is telling us that p. Conceived of thusly, testimonial evidence is tightly connected to both justification and knowledge. By the non-skeptical constraint, we are supposed to have testimonial knowledge about a great many propositions. In what follows, I will take it that an individual H having testimonial evidence for p is a necessary condition on both H having testimonial justification for believing p and H having testimonial knowledge that p. Thus, if a view of testimony entails that we lack testimonial evidence in enough cases, it will entail that we lack testimonial knowledge in those cases, and may thereby violate the non-skeptical constraint.

So, what does it take to have testimonial evidence for some p? As it happens, it doesn't require that anyone testify. Take my counterpart that, rather than getting up, getting dressed and getting to work on this paper, woke to find himself cunningly envatted in a mad scientist's lab, experiencing a simulation of the very things I am experiencing. When the barista tells me that they have no more coffee grounds (which I want for my garden), I get testimonial evidence that the coffee shop is out of grounds. Likewise for my envatted counterpart. On the basis of his becoming justified in believing that the barista is telling him that the shop is out of grounds, he may gain testimonial evidence for the proposition that the shop is out of grounds. The point here is that both my envatted counterpart and I get testimonial evidence, even though only one of us was actually told that p. It sufficed for my counterpart that he was justified in believing that someone testified to him that p. We need not travel to such exotic possible worlds for examples of testimonial evidence that do not depend on anyone actually testifying. All we need is some misleading evidence to justify us in believing that someone is testifying that p when they are not. For example, suppose that Tim is on his way to a meeting, and in passing an open door, sees a professor speaking loudly to a student: ‘The Dean has been arrested on suspicion of embezzlement!’ The professor appears to be in earnest and the student appears genuinely surprised. Suppose, as seems plausible, Tim is justified in believing that the professor is testifying that the Dean has been arrested on suspicion of embezzlement. And on the basis of this justification, Tim may come to have some testimonial evidence that may even justify him in believing the content of what the professor said. However, suppose that Tim has received doubly misleading evidence: he has not just received testimony, the professor and the student are acting out a scene of a play and the Dean has not been arrested for embezzlement. This does not undermine the fact that any evidence Tim thereby received about the Dean's supposed arrest is testimonial evidence. So, for ordinary knowers, at least, it is not necessary that someone actually tells us that p in order to have some testimonial evidence for believing p. In at least some cases, it will suffice that we have justification for believing that someone has testified to us that p.

I will take it that this lesson is general: rather than a particular testimonial speech-act about p being essential to an individual having some testimonial evidence supporting p, what is essential is that one is justified in believing that someone has testified about p. This requirement is neutral with respect to the question of whether reductionism or non-reductionism is true. So, let's define testimonial non-reductionism in the following way:

There are a number of ways of distinguishing reductionism and non-reductionism in the epistemology of testimony. In what follows, the difference between non-reductionism and reductionism is that while non-reductionism requires no further evidence for H to possess testimonial evidence for believing p beyond H being justified in believing that S testified that p, according to reductionism, more evidence is required for H to possess testimonial evidence for p.Footnote ³ This expresses what I take Matthew Weiner to mean by claiming that testimony confers ‘default justification’ (2003: 257). Sometimes too, it is claimed that testimony has ‘a priori authority’ such that it is an a priori truth that ‘beliefs based on testimony are prima facie justified’ (Audi Reference Audi2004: 18). TNR entails this view, at least if it is taken to be the claim that no empirical evidence is needed as positive reasons for thinking that S testifies that p is evidence of the truth of p. Non-reductionists differ about how strongly just having someone's word that p supports p.Footnote ⁴ TNR is consistent with the possibility that one receives very strong evidential support from S testifies that p or very weak support.

According to reductionism, it is not sufficient for H to have testimonial evidence for p that H is prima facie justified in believing that S testifies that p. H must have some additional non-testimonial positive reasons for thinking that S testifying that p is evidence of, reliably indicates that or in some way or another counts in favor of the content of the testimony. Thus, I will define testimonial reductionism as follows:

(b) is intended to capture reductionism's requirement that non-testimonial positive reasons are necessary for H to have testimonial evidence for p. The justification that H must possess is stated somewhat loosely – reductionism requires that there be some sort of reasons that H has that connect S testifies that p and p. S needs to be justified in believing that S testifies that p indicates, is linked to, makes probable or is evidence of p.

In what sense must the justification H possesses be non-testimonial? There are two different readings in the literature. Elizabeth Fricker defends a view she calls ‘local reductionism’ (Fricker Reference Fricker1995: 404). For TR to express something similar to her view, reductionism only requires that one's justification for believing that S testifies that p is an indication that p not depend on this case of (putative) testimony. It remains possible that the justification H has that S testifies that p is indicative of p depends on other cases of testimony. But (b) is also consistent with a stronger reading, that for H to have non-testimonial justification of the sort (b) requires, H's justification needs to not depend justificationally on any testimonial evidence. This is closer to the classical Humean view – testimony really reduces to non-testimonial justification. This ‘global reductionist’ reading of TR is more difficult to defend, but if I am able to defend it, the local reductionist view will also be vindicated. Thus, when I am speaking of TR in what follows, I will intend the stronger ‘global reductionist’ reading.

Detractors of this global reductionist reading of TR have urged that it is liable to a withering objection, which I am calling the ‘NEEO’. A number of authors have made this argument (or close relatives of the argument) including Anscombe Reference Anscombe and Delaney(1979: 143 and 144), Price Reference Price(1967: 117 and 118), Coady Reference Coady(1992: 82) and Lackey Reference Lackey, Lackey and Sosa(2006: 161). I will focus on Coady's version of the objection for a number of reasons. First, his presentation of the NEEO is made in terms of evidence (rather than other epistemic properties) which makes it directly applicable to TR.Footnote ⁵ Second, Coady is particularly influential in shaping the contemporary work on the epistemology of testimony and it is his version of the objection that I see cited most frequently. Lastly, his presentation is somewhat better developed than many other versions of the argument. Coady's target in his presentation of the NEEO is intended to represent a basic form of Humean reductionism inspired by statements like the following by Hume in the Enquiry, in which he claims that our ‘assurance’ in testimony ‘is derived from no other principle than our observation of the veracity of human testimony, and of the usual conformity of facts to the reports of witnesses’ (Reference Hume and Steinberg1993: 94). From this, Coady extracts the following reductionist principle:Footnote ⁶

This Humean principle is congenial to TR, but is narrower. While TR requires non-testimonial justification for believing what we are told in order for a hearer to have testimonial evidence, this principle specifies that these non-testimonial positive reasons must be in the form of observations that constitute grounds for a straightforward induction from ‘a constant and regular conjunction’ between reports and the reports turning out to be true. Of this version of reductionism, Coady writes,

Coady's objection is that we lack samples of testimony and observations that this testimony has turned out to be true that are widespread and varied enough to satisfy the Humean requirement that we have grounds for an induction from observations of reports and the world. If Coady is right, the Humean principle places requirements on our non-testimonial grounds that they cannot satisfy to yield the sort of testimonial evidence we take ourselves to have. Given that Coady is operating with a similar non-skeptical constraint to the one adopted above, this constitutes a good reason to reject the Humean principle above. This is the basic idea behind the NEEO: the actual scarcity of non-testimonial justification we possess, together with a non-skeptical constraint, yields the conclusion that reductionism is false. In presenting his objection, Coady mostly focuses on the Humean principle given above. However, Coady's objection is frequently cited approvingly as a more general objection to principles like TR. Weiner is one such author, expanding on Coady, making it clear that he takes this to be a general problem for views like TR:

The idea is that any form of global reductionism faces the problem of finding non-testimonial grounds for believing that S’s testimony that p counts in favor of p. This is the NEEO:

1. (1) If the non-skeptical constraint holds and TR is true, for most of those propositions p that we take ourselves to testimonially know, we have prima facie non-testimonial justification for believing that some speaker S testifying that p counts in favor of p.

2. (2) It is not the case that, for most of those propositions p that we take ourselves to testimonially know, we have prima facie non-testimonial justification for believing that some speaker S testifying that p counts in favor of p.

3. (3) The non-skeptical constraint holds.

   So,

4. (4) TR is false.

If we assume TR and the non-skeptical constraint, in order for us to have testimonial evidence in enough cases to not violate the non-skeptical constraint, for most of our testimonial knowledge, we must have testimonial evidence. To have testimonial evidence for those propositions p, if TR is true, we must have prima facie justification for believing, in each case of testimonial knowledge, that the speaker S's testimony that p counts in favor of p. Thus, the first premise is true. The second premise, as Coady affirms it above, seems to be an empirical observation – it appears that Coady simply thinks that this is obvious. Finally, the third premise is an assumption introduced early in this section, that the non-skeptical constraint holds.

What's a reductionist to do? I'll concede and resist. I'll concede that Coady is probably right about the Humean suggestion that we have testimonial evidence because we've verified enough testimony to justify a straightforward inductive inference. I will resist, however, the conclusion that premise 2 is therefore true. If other strategies are available to get the kind of prima facie justification for connecting speakers’ testimony with truth of the content of their testimony, then the reductionist can show that premise 2 is false and the NEEO unsound.

2

I have stated that I will draw from the literature on peer disagreement to answer the NEEO. In arguing that discovering that a peer disagrees with us about some proposition p is a reason to revise our strength of belief in p, Richard Feldman writes:

Consider one of the cases of disagreement that Feldman poses:

Suppose that you and I are in the scenario that Feldman describes. In the first stage, we both look into the quad. I seem to see a person in a blue coat and believe accordingly. You seem not to and believe accordingly. Before our disagreement is disclosed to each other, we may both be justified in our respective beliefs. Suppose further that we are justified in taking each other to be epistemic peers: prior to finding out about the disagreement, we have good reason to think that each of us is in just as good an epistemic position as the other to know whether or not there is a person in a blue coat in the quad. Perceptually, intellectually and evidentially, we have no reason to expect that the other is more likely to make an error about this proposition.

The evidential situation changes when we find out that we disagree. Suppose this disclosure happens by us telling one another what we believe and reporting our experiences. Suppose we report our experiences as well. I take it that it is clear in this case that the rational thing for both of us to do is to decrease our confidence in our respective attitudes toward the proposition that there is someone in a blue coat in the quad. Why? As Feldman says, ‘evidence of evidence is evidence’. Finding out that you have evidence that supports your belief that there isn't anyone in the quad in a blue coat is evidence for me that there isn't anyone in the quad wearing a blue coat.

The evidence of evidence principle gives intuitive results in areas beyond peer disagreement. Consider the following sort of case. I once learned a proof for p. I can no longer recall all of the lines of the proof, though. But I remember that I did work through the proof at one time, and that I ‘saw’ that the proof was valid and that each premise was true. The evidence of evidence principle yields the result that I have some evidence for p, because I remember that there was evidence for p. This seems the intuitive answer in this case.

Consider another case. I'm flipping through a colleague's manuscript. I find references to an argument for a conclusion q, but cannot find the argument. Soon, I realize that there is a page missing right about where the argument should be. If I know that she is unlikely to write a bad argument, I have, even without having the argument before me, some reason to think that q is likely to be true because I have evidence of her evidence for q. Again, I find this the intuitive judgment on the case. Thus, there is at least some defeasible evidence that the evidence of evidence principle is true.

Given this brief defense of the evidence of evidence principle, here is the principle more precisely, which I will take to be true in the rest of this paper,

A note about EE is in order. EE does not entail that x, who has some evidence e that y has some evidence é that supports p, is thereby justified in believing p. For one thing, x might have some further evidence é´ that is a defeater for x, so that either e's support for p is either swamped by evidence that supports not-p or that undercuts x's evidence e for p. EE is also consistent with e only supporting p to a very weak degree. But even in this weak form, EE is sufficient for my modest purposes below.

3

If we typically had evidence that what people are testifying to us is accompanied by evidence, then we would have a way of rebutting the second premise of the NEEO. My first step in showing that we do typically have such evidence is to assume some general principles about conversational norms. Consider H. P. Grice's Cooperative Principle (hereafter ‘CP’),

I'll assume that the CP holds for those informational talk exchanges in which testimony occurs.Footnote ¹⁰ I will assume also that, as competent language users, we typically have prima facie justification for believing that we are in informational talk exchanges when we are conversing.

When Grice introduces the CP, he introduces a host of maxims he seems to think follow from the CP. For example, the maxim of quality,

If one is to follow this maxim, the means of doing so are clear: one ought to assert what is supported sufficiently by one's evidence, which Grice codifies in a sub-maxim, ‘Do not say that for which you lack adequate evidence’ (1978: 46). Grice's wording of this sub-maxim is unhelpful. Adequate evidence for what? If we rely on evidentialism about justified belief here, there will be a degree of evidential support that will be sufficient for belief (Conee and Feldman Reference Conee, Feldman and Smith2008: 83). A natural degree of evidential strength to require would be that which is sufficient to justify belief. So, I will assume that in all cases where the CP holds, the following clarification of Grice's sub-maxim also holds:

EMQ is an imperative. People may or may not follow it. But, if we had reason to think that they did, we would have moved further forward toward showing premise 2 to be false. If we were justified in believing that people successfully follow EMQ, then we would have evidence that people testify only to what they have sufficient evidence for believing. By EE, we would thereby have evidence of evidence in all of those cases of testimony.

But do we have good reason to think that people follow EMQ? I will offer two arguments for the conclusion that we do.

My first argument extends the Gricean picture assumed above. Conversational implicature is closely related to these rules and maxims. We routinely express more by our utterances than the strict lexical meaning of our sentences. In Grice's words,

In the above, Grice is concerned with the conditions for the speech act of implicating some proposition, rather than with epistemic conditions for being justified in believing the implicated proposition. But, if we are ever to be justified in believing that some speaker is implicating a proposition like q, the presumption Grice mentions above that the speaker S is following the maxims must at least include prima facie justification for H in believing that S is observing the conversational maxims. If the presumption is a mere assumption, then as the hearers of implication, we would never be more than conditionally justified in believing that someone is implicating a proposition like q when they say p. Thus, if we are be justified in believing that someone is communicating q to us when they say p (thereby implicating that q), we must in general have justification for believing that speakers are successfully following CP and the conversational maxims (including EMQ). Otherwise, we are only justified in believing that they are implicating q by saying p if they are following CP and the conversational maxims.

Here it is worth considering a potential objection.Footnote ¹² One might worry that my requirement here is too strong; perhaps, in order to be justified in believing that someone is implicating q by saying p, we only need to be justified in believing that they are sincere, that they are trying to follow the CP and EMQ. If EMQ is something that we are only justified in believing that speakers try to follow, then we do not have evidence of evidence. I suspect that success in following the CP and EMQ is needed for being epistemically justified in believing that a speaker implicates that q by saying p, rather than mere sincerity. However, here I will defend a weaker claim, that typical hearers are justified in believing that if people try to follow EMQ, they will successfully follow EMQ. There are two ways in which a person might try to follow EMQ and fail to: mis-speaking and failure to correctly assess what one's evidence supports. I will argue that hearers will usually be in a position to dismiss each of these two possibilities and in those rarer cases where a hearer is not, these cases do not present a problem for my defense of TR.

The first of these cases are ones in which S might be justified in believing p and try to testify that p, but fail to testify that p. This is mis-speaking: intending to communicate p, but communicating some other proposition q unintentionally. The cause could be linguistic incompetence, perhaps a person is just learning the language in which he or she is communicating and chooses the wrong expression for p. Often it will be just a mis-speaking: saying a word one did not intend to say or unintentionally picking a word that is ambiguous in a context. When we detect a high risk of mis-speaking, we should have less confidence in believing what it seems to us is being testified about. But we can usually dismiss this possibility in conversation because speakers will often correct themselves, conversational context will often allow us to understand what someone meant despite what they said, and we will often have strong indications of linguistic incompetence when we are conversing with or reading something written by a speaker struggling with our language and conversational conventions.

More troubling for my argument is the possibility that S does not mis-speak in testifying that p and has misjudged that p is the proposition that S's evidence justifies for belief. If this possibility cannot be dismissed, then we do not get evidence of evidence from testimony, but only evidence that someone tried to testify to what their evidence supports. Unlike the case of mis-speaking, where when a person is a competent language user it is likely that they will be in a position to detect that something has gone wrong, this scenario is a skeptical one. It seems or appears to S that p is justified for her belief, but p is not. I think that hearers are usually in a position to dismiss this skeptical possibility. I fear, however, that defending this view in detail will take us too far afield. I will instead suggest two ways of arguing that two independently plausible epistemological views both have the result that if S tries to follow EMQ, then, unless S mis-speaks, S will succeed in following EMQ.

The first thought is that an additional higher-level justification principle will help. Suppose that if S is justified in believing that S is justified in believing p, then S is justified in believing p.Footnote ¹³ Then, when S tries to testify that p (because S is justified in believing that S is justified in believing p), S will successfully comply with EMQ (so long as S does not mis-speak).

The other solution appeals to phenomenal conservatism. It is plausible, that if S tries to follow EMQ, S will testify that p only if p seems to be justified for S. Suppose that a seeming that q confers prima facie justification on q and thereby is evidence to some degree in favor of q (relying on my evidentialist assumption above).Footnote ¹⁴ It will follow then that when S tries to follow EMQ in testifying that p, S is prima facie justified in believing that S is justified in believing p. Applying the evidentialist assumption, H possesses evidence that S possesses evidence that on balance supports p. By EE, S possesses evidence that p when S tries to follow EMQ. A full defense of my view would require defending these principles, but doing so is beyond the scope of this paper.Footnote ¹⁵

If I have succeeded in the foregoing, for all hearers H who are able to recognize putative cases of implicature in a language L (that is, are in a position to be justified in believing that S communicates q by saying p) and are justified in believing that some speaker S is speaking in L, then H will be justified in believing that S is following CP and the maxims (including EMQ). Let the class of all such hearers be C. The following principle will be true in the class C:

For hearers in C, whenever any of them have prima facie justification for believing that some S testifies that p, then that hearer H has prima facie justification that S is justified in believing p. EET being true in C provides a way for the defender of TR to respond to the NEEO. According to TR, in addition to having justification for believing that some S testifies that p, a hearer H must also have justification for believing that S testifies that p is an indication of the truth of p. Can we show, given the above, that this is true, at least in a sufficient class of cases for satisfying the non-skeptical constraint? I think so. First, I'll show that for hearers of testimony in C, it will follow that these hearers of testimony have testimonial evidence according to TR. I will then argue that they are in a class of hearers sufficient to satisfy the non-skeptical constraint.

I'll need to show that for these hearers of testimony in C, if they have prima facie justification for believing that some S testifies that p, then they have prima facie justification for believing that S testifying that p counts in favor of p. Suppose that H is a hearer in C. Suppose further that H is justified in believing that some S testifies that p. By EET, it follows from that supposition that H is prima facie justified in believing that S is justified in believing p. I will rely on evidentialism here. I will assume that if any S is justified in believing p, then S has a body of evidence that on balance supports p (Conee and Feldman Reference Conee, Feldman and Smith2008: 83). If so, S being justified in believing p entails S having a body of evidence that supports p. So, H has evidence that S has evidence that supports p and by EE, it follows that H thereby has (some) evidence that supports p. Thus, it follows that whenever H has prima facie justification for believing that some S testifies that p, H has evidence that S's testimony that p counts in favor of p, because of the truth of EET in C.

Thus, it follows that in C, these hearers will satisfy both (a) and (b) of TR. That is, in those situations where some hearer H in that class is prima facie justified in believing that some S testifies that p, H will also be justified in believing that S testifying that p counts in favor of p.

I now wish to argue that C is sufficiently large for TR to satisfy the non-skeptical constraint and to thereby show premise 2 of the NEEO to be false. Note that most of us, when we are hearers of testimony, are usually in a situation where we are justified in believing that we're in a situation where a language we're a competent user of is in use and we are prima facie justified in believing that those we're listening to are complying with CP and EMQ. Thus, in most situations in which we receive testimony, we satisfy the requirements of TR for having testimonial evidence for the content of what we are justified in believing that our testimonial partners are saying. So, widespread testimonial skepticism will not follow if TR is true.

C is hardly an elite class; most of us are members of C in our communicative exchanges. It will be instructive to note some ways in which hearers will not be members of C. Among other things, it will allow me to distinguish the defense of TR against the NEEO from defending TR against another related objection. Sometimes hearers are not justified in believing that a competent linguistic exchange is going on, either because of problems on the hearer's side or problems on the speaker's side. One way this may happen is when the hearer is not a competent speaker of the language of the communicative exchange. If the hearer is not in a position to recognize putative implicature, then they will not count as a member of C. But, it is clear that the non-skeptical constraint introduced earlier should not include these cases as a constraint on the epistemology of testimony.

Problems may occur on the speaker's side too. If I am in a conversation with someone who is not a competent speaker of English (but is a competent speaker of another language), I will have reason to wonder if the sentences uttered really mean what the speaker is trying to say. This may not leave me out of class C above, because I may be able to recognize putative implicatures during the exchange. Thus, I may receive evidence of evidence in such exchanges, but I will also receive defeating evidence, precisely because I will have reason to doubt that the individual is successfully following CP and the maxims (even if they are trying very hard to do so, but just don't have the linguistic skills to do so). A defense of TR should not leave us in a very strong epistemic situation in these cases. My defense does not entail that we are.

The testimony of children and the status of the testimony that young children receive are of interest here also. When an adult receives testimony from a child, we may have room to worry that the child is saying what they mean to say. They are not yet a competent speaker of our language, after all. So, though in some cases, we will be in class C with respect to children, we will have good reason to suspect that we might be mislead by a speaker who simply is having a hard time making himself understood. In other cases, we will have good reason to doubt that a child is following CP and the maxims.

What about children as hearers of testimony? Children as hearers of testimony are sometimes supposed to pose a problem for reductionism called the ‘infant-child objection’. The infant-child objection is related to the NEEO, because it also starts with a kind of non-skeptical constraint and argues that reductionism is not able to satisfy it. The defense I outline here against the NEEO will yield the conclusion that children have testimonial evidence according to TR so long as they are members of C.

It is less clear that very young children will be members of C. On the view I have outlined above, a child is a member of C if they are in a position to be justified in believing that someone is communicating p by implicature by saying q. Though I suspect that many children are members of C quite early, I do not wish to argue this here, because I am not trying to answer the infant-child objection, which is a matter for another inquiry. Instead, my argument here defends TR against the NEEO, where the key threat is that obvious cases of testimonial knowledge, of the sort enjoyed by mature speakers and hearers, seem to be threatened by TR.

There is another independent argument for the conclusion that premise 2 of the NEEO is false: an argument from defeat. Consider a case. Smith sees a cardinal in a bush and comes to believe on the basis of this seeing that there is a cardinal in the bush at a time t. Suppose that this constitutes justifying evidence for his belief that there is a cardinal in the bush. Suppose that five minutes after his experience and believing the cardinal proposition, Smith receives excellent evidence that he has been hijacked and immersed in an illusory sensory-simulation for the last ten minutes. Suppose, for the purpose of the argument, that he has evidence that justifies his believing with certainty that he has been in this sort of simulator. What is the status of his belief that there is a cardinal in the bush? Any support that his experience conferred upon the cardinal proposition has been completely nullified, because he now has conclusive reason to doubt that his experience was veridical. Call this kind of defeater for the cardinal proposition's support from the appearance as of seeing a cardinal in the bush a ‘full-defeater’. Because all of Smith's support for the cardinal proposition was from that appearance as of seeing a cardinal in the bush, getting justification to the point of certainty that he was in a sensory-simulator completely nullifies Smith's support for the cardinal proposition.

Consider a second case in which Jones tells Smith that there is a cardinal in the bush at t and on the basis of this testimony, Smith believes that the cardinal proposition is true. Suppose further that Smith has testimonial justification for believing the cardinal proposition. Shortly thereafter, at t′, Smith gets justification to the point of certainty for believing that Jones had no evidence about whether or not there is a cardinal in the bush. What is the status of Smith's belief that the cardinal proposition is true? It seems that Smith has received a full-defeater for the cardinal proposition at t′. Coming to have very strong justification for believing that Jones had no evidence for the cardinal proposition fully defeats Smith's justification for believing it. The obvious explanation for this is that Smith's having justification for believing the cardinal proposition depended upon Smith possessing evidence of evidence, specifically, that Jones had evidence that supported believing the cardinal proposition.

But is this the best explanation of the full defeat that Smith receives by becoming so strongly justified in believing that Jones has no evidence about whether or not there is a cardinal in the bush? If a good non-reductionist solution is available, this would undermine my case. One explanation that might be attractive to non-reductionists seems unlikely to work here. Non-reductionist accounts of testimonial justification typically will include a no-defeaters condition: testimony justifies belief in its content so long as we don't receive a defeater. This coheres well with the account of non-reductionism about testimonial evidence offered above as TNR. According to TNR, having reason to believe someone is testifying that p is itself evidence that p is true. Evidence that p is true will only be sufficient for justification for p if one does not have any defeaters that prevent one's testimonial evidence from justifying p.

For example, if Jan finds out that Jim is a habitual liar, she comes to have some reason not to believe Jim's claim to have completed a solo free climb of El Capitan last weekend. Taken by itself, as TNR claims, Jim's testimony still is evidence for Jan that he solo free-climbed El Capitan last weekend, but Jan's evidence, taken as a whole, no longer supports that he did to the same degree (if at all). So, could a non-reductionist argue that something similar is going on in the case of Smith, Jones and the cardinal in the bush? I don't think so. Here is why. If we find out that someone is a liar, then we come to have some reason to think that his testimony is false. So, this likelihood of falsity should be weighed against whatever the initial support the content of the testimony receives (from TNR) and this will determine the overall support the content of the testimony for an individual. But consider the proposition Jones had no evidence about whether there was a cardinal in the bush. This does not directly bear on the truth of the proposition there was a cardinal in the bush. So, for the non-reductionist, there will have to be some additional proposition that, together with the proposition about Jones's evidence, will account for the full-defeat. But the likely candidates will not explain full-defeat of Smith's testimonial evidence, only partial defeat. A clear candidate for linking the proposition about Jones's evidence with the cardinal proposition would be testifiers who do not have evidence supporting the proposition they testify about are unlikely to testify the truth. But note that this linking proposition will only be to some degree justified for Smith, not justified to the point of certainty. Thus, it will remain to some degree likely for Smith that Jones is telling the truth, and thus, with this linking proposition, the non-reductionist account does not predict full defeat for the cardinal proposition, only partial defeat. Thus, the obvious non-reductionist explanation of why it is that finding out that a testifier does not have evidence will not predict full defeat in the case of Smith and Jones above, only partial defeat. Thus, unless a better explanation can be offered by non-reductionists for why it seems that Smith's testimonial justification for the cardinal proposition is fully defeated in this case, it would appear that my reductionist explanation, that Smith's having justification for believing the cardinal proposition depended upon Smith possessing evidence of evidence, best explains our intuitions.

If this is right, this conclusion generalizes to all cases of testimonial evidence (where this class is as widely defined as our pre-theoretical intuitions dictate). For any individual H with testimonial evidence for some proposition p, it will suffice to fully defeat H's testimonial justification for believing p for H to become justified to the degree of certainty that the testifier S had no evidence for p. Thus, we have reason to think that in any case where some H has testimonial evidence for some p, H has evidence that the testifier S had evidence that p. This is equivalent to EET being true in the class of all individuals with testimonial evidence for any proposition.

Thus, this argument from defeat suggests a stronger conclusion than the first argument. Instead of EET being true in a restricted class of hearers of apparent testimony, this is an argument that there is good reason to think that EET is true in all cases where some H has testimonial evidence. For the same reasons offered in the first argument against premise 2 of the NEEO, if EET is true, there is evidence that any S such that an individual H in that class is prima facie justified in believing that S testifies that p, has a body of evidence that supports p. Again, according to the line of reasoning offered above, H will therefore have prima facie justification for believing that S testifies p is an indication that p. Thus, for the class of all individuals in possession of testimonial evidence (a broader class of hearers than the non-skeptical constraint requires), TR is satisfied, and premise 2 of the NEEO is false.

4

The NEEO fails. I have offered, in the end, a modest defense of a modest form of testimonial reductionism. My argument is not sufficient to show that TR is true, only that TR is not shown to be false. It is not sufficient to show that the evidence we get in testimony via EET and EE is evidence that justifies belief strongly enough for knowledge. These are issues that deserve further exploration, but which I will not take up here.Footnote ¹⁶