Have intuitions, will travel — but not very far this time without your help. Recent work by Peter Klein on benign falsehoods leads to the recognition of conflicting intuitions about testimonial knowledge in a particularly bewildering case for all parties in the epistemology of testimony. In what follows, I sort through this thicket of problems with a view to laying out the hard choices. In the course of doing so, I also make some exploratory remarks for an epistemology of fiction.
1. Little Virginia, Santa Claus and testimonial justification
Consider the elements of the problem. We’ve been discussing how best to account for the apparent acquisition of inferential knowledge from false premises (that is, false beliefs on which a given belief is apparently both causally and evidentially based). The details of that discussion will not occupy us here.  Suffice it to note that there are cases of inferential knowledge from beliefs which are false, testimonially justified (or warranted, whatever the preferred analysis of the concept), and epistemically benign, that is, are contextually such that certain true beliefs based on them clearly seem to be cases of inferential knowledge. Here is one such case:
The Spokesperson: The very reliable spokesperson for the president assures me that (q) the president is in Jordan. Based on my belief that q, I infer that (p) the president is not in the Oval Office. Suppose that p is true and q is false: A last-minute change in his Middle Eastern tour schedule now has the president in Israel, not Jordan. Don’t I know that p? 
This is a case of inferential knowledge from a benign false belief that is testimonially justified. But there is nothing here that would be of any use for the epistemology of testimony. Now, in setting up his analysis of the benign falsehood phenomenon, Klein (2007) puts forward this case of putative inferential knowledge:
The Santa Claus Case: Mom and Dad tell young Virginia that Santa will put some presents under the tree on Christmas Eve. Believing what her parents told her, she infers that there will be presents under the tree on Christmas morning. She knows that.
I have agreed with Klein that this is a case of inferential knowledge. I have also put it to the test a number of times in talks on the issue, and my audiences invariably accept it as a case of inferential knowledge. But this is a putative case of knowledge from testimonially justified false belief that does seem to wreak havoc with some of the most important intuitions in play in the current debate in the epistemology of testimony. The problem is brought to my attention by Felipe Müller’s question (in conversation): How can little Virginia’s premise-belief (that Santa will bring presents) be testimonially justified?
2. Credulism, epistemic growth and testimonial knowledge
The obvious answer to that question is that there is a seemingly tenable externalist thesis in the epistemology of testimony according to which Virginia’s premise-belief does acquire positive epistemic status (justification, warrant…) from the reliability of her sources. In Duncan Pritchard’s (2004) words, the externalist thesis that concerns us here is the following:
Bare Credulism: For all of one’s [testimony-based beliefs], one’s TBB can be justified even though one is unable to offer any supporting grounds in favour of that TBB. 
The obvious answer must be wrong. Some of its implausibility comes from the troubles of anti-reductionism in the epistemology of testimony. Why would anyone want to defend any form of Credulism in view of its apparent absurdity?  The pre-theoretical answer appeals to putative counterintuitions in everyday cases where testimony is ostensibly accepted in the absence of justifying grounds: We do solicit information from strangers on a regular basis (“which way to x?”, “what time is it?”, “do you know whether x?”…), and there are those who think that that would hardly make any sense unless we were inclined to believe what we are told. But this is a flimsy case at best. We also have a very substantial intuition according to which the usefulness of such everyday offer of information does not depend on our believing our sources. If the stranger tells me that my destination will be found around the corner, and I have no practical way of checking for reliability, I give it a shot and look around the corner. The situation is akin to betting (regardless of the existence of beliefs about the odds): I have little to lose and much to gain from acting as if I had been told the truth, though actually believing what I have been told by a complete stranger is no indispensable part of an appropriate response to the offer of information. Intuitions are wobbly at best. Sometimes it looks like there’s belief there; sometimes, it doesn’t. My undergraduate students are split on this. (Some think they always believe; some think they never do; and there are those who are simply perplexed by why they sometimes do and sometimes don’t – and why, in any case, it could ever be reasonable to do so.) There is no solid pre-theoretical case for Credulism. 
We then come to the theoretical case. It’s built around the claim that some form of Credulism is needed to account for epistemic growth. In Elizabeth Fricker’s (2004) words,
we cannot explain the status as justified and knowledgeable of a person’s system of beliefs without invoking a [credulist thesis] applicable at least in her developmental phase. But this does not entail that for a mature individual, now possessed of a commonsense view of the world and her place in it, this epistemic right ungroundedly to trust still holds. One way of explaining such a distinction between the developmental phase and the mature phase is to hold that an infant starts off endowed with the epistemic right to trust what she is told as such; but by the time she has matured, this has been permanently defeated by the knowledge she has acquired of the deceitfulness and folly of human nature: once one is old enough to know better, one should not trust blindly what others tell one. 
The question that this immediately elicits is: Why is the positing of an epistemic right of which we have no independent evidence (and would deny for mature adults) be the best explanation of what goes on in a person’s developmental phase with regard to her believing what she is told on blind trust? The question is urged by the apparent availability of this alternative explanation of a small child’s exercise of blind trust: While the child’s reaction to testimony can only be one of blind trust, she can have no testimonial knowledge. The best she can get (and does get) from reliable testimony is true belief. Why is that a problem? Although she doesn’t explicitly consider the alternative, we find this answer in Fricker’s work:
If there can be no thought without language, and if — as is almost certainly a law of human psychology and neurobiology, even if not logically necessary — a human being cannot acquire language except through being taught it, then it is at least a psychological necessity that all humans have extensive historical dependence on testimony in their development of a system of empirical belief.
This is her basis for concluding that
the prospects of eliminating dependence on ungrounded trust are as we saw: it cannot be done, since the very idea of setting aside all my beliefs which depend on past simply-trusted testimony, to reconstruct my entitlement to them out of the materials that are left, is absurd.
But what is the basis for thinking that what is acquired at the developmental phase (concepts, language, the barest commonsense view of the world) should count as propositional knowledge and not mere true belief (which is clearly sufficient for all practical purposes)? If I’m not mistaken, this is as close as Fricker comes to facing that hard question:
The early phases of language-learning involve such accepting reactions to ostensive teaching of word meanings. Quibbling over whether this is precisely ‘testimony’ does not do away with the basic fact that an infant’s learning of language involves unquestioning practical acceptance as true of what her teachers say to her.
No, let us not quibble over whether what the child feeds on can properly be called “testimony”. I will be happy to call it “testimony”. The hard question that is dodged here (by every author in the field, as far as I have been able to determine) is whether a child’s acceptance of veridical testimony, of propositional knowledge that is conveyed by trustworthy testimony, on the basis of blind trust can properly be called “testimonial knowledge”, whether that is the best we can do with the data.
We have an alternative. Suppose that, at a given moment, all that a very small, blindly trusting child can get from testimony is only true belief, precisely because, as far as testimony is concerned, she’s essentially on a par with the cognitively impaired (mentally retarded, psychotic, aphasic) adult to whom we would unhesitatingly deny entitlement (warrant, justification…). When the child acquires the concepts required for judgments of reliability, all the veridical testimonial beliefs that persist through the move into a phase of higher cognitive competence are immediately turned into knowledge en bloc. (A detailed description of the change in epistemic status may be complex, requiring the satisfaction of some deliberation requirement, or maybe just the disposition to trust on the basis of grounds for thinking the testifier reliable, etc. But this should suffice for our purposes.) The massive change in epistemic status of the child’s veridical TBB’s may be effected all at once, as a result of the child’s keeping all of those beliefs acquired from sources that she now (somehow) acknowledges as reliable. (Again, I’m not trying to describe the process in any detail.)  Is this absurd?
If it is, Fricker does not have a convincing explanation of why it is. And I can’t see a defensible positive answer to that question that would not ultimately rely on linguistic evidence from everyday, ordinary knowledge attributions to children, animals, and even inanimate objects (“the elevator door knows when to close”, etc.). But that kind of evidence is notoriously problematic. We know that ordinary attributions of knowledge are routinely retracted when the attributor becomes aware (is reminded) of the distinction between knowledge and true belief. It happens in our undergraduate classes all the time! It is a well-known fact that, in many contexts, all intended purposes are well-served by calling “knowledge” what we would more cautiously refer to as “true belief” in contexts where the distinction matters. Even philosophers who are well-aware of the distinction — even those whose awareness of the distinction is beyond suspicion — occasionally find no harm in calling “knowledge” what they know is only true belief. Thus, Alvin Goldman (1999):
By know, in this context, I mean only: have an accurate, or true, belief. I do not mean: have a justified true belief (or whatever else is entailed by the richer concept of knowledge).
How reasonable is it to expect the non-philosophical speaker to make the distinction in ordinary parlance?
But, if it is not this kind of assumption — the perceived need to render an analysis of TBB’s that accounts for cognitive growth with knowledge at the base, with knowledge playing the role of cognitive propeller — that is tacitly operative in Fricker’s account (her motivation to insist that we must account for the acquisition of testimonial knowledge at the developmental stage), I can’t see what is. The mistake is to think that it is impossible for one’s true TBB’s that survive some reliability check (that I’m not concerned with describing) to turn into testimonial knowledge all at once. This will, indeed, be impossible if the acquisition of concepts and language must itself occur only on the basis of (a posteriori) propositional knowledge. But there seems to be no reason to think that that is the case. As Robert Audi (2003, 143-144) notes,
[a]t the time concepts are initially grasped in childhood, it may not be necessary that (propositional) belief and knowledge are acquired in every case. Conditions sufficient for conceptual learning may not be automatically sufficient for propositional learning… Testimony…can be concept-producing, belief-producing, or both. The former case seems to be more primitive, and the conditions for its possibility should not be taken as sufficient for the possibility of the latter. 
3. Lies, testimonial knowledge and benign falsehoods
It is noteworthy that the Santa Claus case — insofar as it is reasonably taken to be a case of inferential knowledge — connects in an interesting way with Jennifer Lackey’s (2007) claim that testimonial knowledge is not essentially knowledge transmitted by testimony, since the testifier may have no knowledge (or justified belief, or even belief) to transmit. One of the two very salient and essential features of the Santa Claus case is that Virginia’s parents are lying to her. And, yet, that apparently does not keep her false TBB (that Santa is coming with the goodies) from being a benign falsehood. Since benign falsehoods are as good a basis as knowledge for the acquisition of inferential knowledge, the Santa Claus case seems to provide evidence for a very obvious consequence of Lackey’s claim: Epistemic good can indeed result from insincere assertion in some contexts.
4. Santa, we hardly know you: notes for an epistemology of fiction
The other very salient and essential feature of the Santa Claus case is what makes Virginia’s false TBB such an interesting benign falsehood: It’s about a fictional character. And that should make us ponder an intriguing possible explanation of the intuitions in play in the case.
David Lewis (1978, 1983) called our attention to the fact that there is such a thing as truth in fiction. (Call it “f-truth”.) It is f-true, he notes (without using the term), that Sherlock Holmes lived at 221B Baker Street, but f-false that he lived in a bank, assuming it true (but f-false) that there was a bank at that address in Victorian London. (What is f-true can be true in the world where the story is told as fiction, provided the f-fact is not nomically or otherwise impossible there. Call this assumption “Coincidence”. More about it shortly.) In the same vein, we should say that it is f-true that Don Quixote was a stubborn man, that Lady Chatterley was no deceiver, that Santa Claus is busy on Christmas Eve. There is, indeed, this intuitive aversion to saying that it is simply false (or neither true nor false) that, for instance, Santa is busy on Christmas Eve. Santa is, of course, busy on Christmas Eve — even to those who believe it false that there is a Santa in our world.  What, to my knowledge, has not been noticed (or analyzed in any detail) is that, if there is f-truth, there must be f-knowledge, and also f-justification (f-warrant, f-entitlement…). We do need such notions to account for the fact that not all of one’s non-true beliefs tell against one’s reliability as a believer (or testifier). From the fact that all of the beliefs that an exclusive reader of fiction forms on the basis of what he reads are false (or neither true nor false), it would be fallacious to conclude that, if he tells you he has read that Don Quixote had a dependable friend, you shouldn’t believe that Don Quixote had a dependable friend. The testifier may be very f-reliable. And f-reliability may be indistinguishable from reliability as a sign of certain desirable traits of epistemic character. There is a crucial difference between your beliefs which are f-justified and the reveries of a lunatic, even if they are the same beliefs (beliefs with the same content). This is more obvious when the beliefs are true: The reveries of a lunatic obviously can’t be knowledge when true – on the commonsensical assumption that mental illness (some form of it, in any case) is a decisive bar to reliable believing. But they can’t be f-knowledge either when f-true. Madness precludes f-knowing too! We shall now look into the matter.
As regards our understanding of the epistemic workings of testimony, the distinction between f-knowledge and knowledge, f-justification and justification seems largely (perhaps entirely) irrelevant. Consider the main features of testimonial f-knowledge.
The more natural — or, in any case, the simplest — account of the structure of f-justification is foundationalist.  There is basic f-knowledge. Testimony by the storyteller is the only source of basic f-knowledge. If you get the story from its author, it’s f-knowledge. Foundational f-justification is infallible. 
But notice how this gives us our Hard Choice #1: The epistemology of fiction does seem to provide us with a cozy theoretical environment for Credulism. On a foundationalist account, we rely on the peculiarities of f-assertion, assertion within a context of storytelling. F-assertion is peculiar in that the ritual that constitutes the act of storytelling does itself, as a rule, establish the storyteller as the sole, infallible authority on the matters reported on in his storytelling. Blindly trusting what the storyteller asserts does seem to be the appropriate response to storytelling. It does seem to be essential to constituting somebody as audience of an act of storytelling. (Think of how bizarre a skeptical attitude would be on the part of the hearer: “Then the starship landed on…Jupiter?! You’ve got to be kidding!”) Which seems to support this form of Credulism in the epistemology of fiction (or “Ci-Fi”, for short): Any TBB caused by f-assertion is f-justified (just by being so caused).  And yet, even here, Credulism may seem unacceptable in the absence of much non-obvious qualification. A crucial aspect of f-justification is that the storyteller – either the author or a suitable representative – must establish his authority in those particular matters of f-truth. He must either submit his credentials as author (and be accepted as such), or must give us some evidence that he has read the original story, or has been told the f-truth by somebody in the f-know, etc. Those of us who are not persuaded by the available case for Credulism (section 2 above), will likely be suspicious of Ci-Fi too. One reason to be suspicious is that it’s hard to see how those who – like very small children and cognitively impaired adults – cannot acknowledge a testifier’s authority on matters of truth might be better-positioned with regard to matters of f-truth. Mustn’t one be able to make the distinction between reliability and f-reliability as a necessary condition of f-knowing? And how about mixed contexts — as when one flips into and out of storytelling repeatedly? It is a staple of comedy (particularly stand-up comedy) that the testifier is not to be trusted entirely; but the comedian does purport to be partially trustworthy. When the partial trustworthiness is not picked up by the audience, the joke may fail.
Our foundationalist account of f-justification does, in any case, give us a rationale for denying that, when we have a belief that can at best be f-true (because, say, it is nomically impossible in the world of the believer), we are like the mad believer who holds the same belief. If the structure of f-justification is foundational, with basic f-knowledge at the base, every f-justified belief is causally traceable to an act of storytelling. No such act of storytelling, we may assume, can be found as a causal basis of the mad belief (except, perhaps, schizophrenic storytelling to oneself). There is no problem that I can see with holding that the mad believer is a prisoner in his fictional world. And that fictional world may conceivably be the world where the action described in our f-justified beliefs takes place too. But our foundationalist assumption about how basic f-knowledge is acquired implies that the f-justified believer’s access to the fictional world is provided by testimony (in the act of storytelling), whereas we may assume that the mad believer’s access to that same world is provided by his illness. 
My story about f-knowledge is a chapter in the epistemology of testimony.  A moment’s reflection suggests that there cannot be anything like perceptual f-knowledge. Suppose that you see a man in a Santa Claus suit robbing a bank. The man looks exactly like Santa, and he looks exactly like he’s robbing a bank. Can you be f-justified in believing that Santa is a bank robber? Of course, not. (Never mind the fact that you may need at least some non-perceptual f-knowledge in order to be described as perceptually f-knowing any f-truths about Santa. This is still as close to a perceptual f-knowledge case as we can get here.)
The reason why you cannot be f-justified in believing that Santa is a bank robber makes us consider inferential f-knowledge. A work of fiction gives you basic f-knowledge in the storyteller’s text/speech. It also constrains the inferences you can f-justifiedly make about its characters. From what you read in Conan Doyle’s stories, you are f-justified in inferring that Holmes’ IQ is higher than Watson’s. You f-know it, though not because you find that assertion in the novels. Likewise, even though you f-know that Santa has a lot of free time in his hands (it’s inferential f-knowledge for you) and you have no basic f-knowledge that he is not a bank robber, you do inferentially f-know that he is no bank robber in his spare time.
This brings us to the social dimension of f-justification.  Suppose you know that you are now looking at an unpublished manuscript by Conan Doyle. It’s another Holmes story. In it, the author claims that Holmes is actually the murderer in some of his most famous cases. He is said to have killed the victims and brilliantly forged the evidence with which he framed the accused. (Suppose the forgery was so brilliant that it instantly drove the accused mad, making them confess to being the murderers they in f-fact were not.) Holmes the murderer in his own cases?! F-false! Not even Conan Doyle had the power to produce such a dramatically surprising turn of f-events. We wouldn’t recognize the Conan Doyle we once knew. We would seriously consider the hypothesis that he had gone mad while writing the sequel, or the hypothesis that the manuscript is just a bad joke, or maybe it’s a draft Conan Doyle himself discarded, or maybe we should think the faux Holmes a different character altogether. The bizarre assertions wouldn’t cause any TBB’s (or, at any rate, wouldn’t f-justify any TBB’s). Lady Chatterley a transvestite?! F-false! Don Quixote an alien from outer space?! F-false! Santa Claus painting the town red with his buddy Count Dracula?! F-false! 
If this is on the right track, the author’s (the original storyteller’s) infallibility vanishes sometime after the original story is told. At some point – a point of which we may remain ignorant – the storyteller’s authority can be reasonably questioned. We’re still getting basic f-knowledge from him (as the story goes on in subsequent episodes of storytelling about the same characters), but it’s f-knowledge resting on fallible foundations. What the storyteller puts forward as testimony will reasonably be questioned – the very identity of the characters will be jeopardized, his testimonial authority will come under suspicion (his sanity, his seriousness, etc.) – when his account of the f-facts becomes wildly incoherent with what we originally f-knew about those characters. But, notice that nothing here conflicts in any obvious way with the storyteller’s presumptive infallibility when the story is originally told. Nor is there any obvious difficulty in our describing this account of the structure of f-knowledge as a form of foundationalism. Recall that there are versions of foundationalism (“moderate foundationalism”, of one kind or another) according to which a belief is epistemically basic iff its prima facie justification does not depend on any beliefs. Ultima facie justification is a different matter altogether.  Foundational beliefs may not be ultima facie warranted when they are incoherent with the doxastic superstructure.
But now we must face our Hard Choice #2: It seems entirely unproblematic to hold that, as regards inferential f-justification, the author of the story is epistemically on a par with his audience: The author may lack f-justification for some of his inferential beliefs because they arise from bad inference from what he f-knows. We have also seen how his would-be basic f-knowledge is constrained (defeasible) by the widespread beliefs that he himself generated with his original story. But where does his basic f-knowledge come from? He has simply made his story up. A tentative answer: The author does have basic f-knowledge and his basic f-knowledge is a priori. It shouldn’t be a problem to think that, if something that you make up is an item of knowledge at all, it is an item of a priori knowledge. What could be more a priori to you than what you yourself have made up?! That surely explains why you are an infallible testifier. The problem is in the idea that anything that can be fabricated at will may deserve to be seen as an instance of the natural kind knowledge. It is entirely not obvious that the author believes what he has made up. On the other hand, there is this seemingly strong intuition that he can believe what he infers from what he has made up. But, again, consider the counterintuition: Apparently, our only hope of distinguishing an act of literary creation from mad believing lies in the claim that, unlike the madman, the author does not believe what he has created. And yet, again: Why not accept that, with regard to an item of putatively basic f-knowledge, the author is indistinguishable from the madman with the same belief? What would then make the desired distinction is the fact that the author is not a prisoner in his fictional world. Unlike the madman, he does go back and forth. We might then try to account for the distinction with the following requirement: A person can f-know only if she can know.
To the best of my understanding, Lewis rejected Coincidence in the following passage of his 1978 paper:
[A] real-life Holmes might have had his real-life Watson who told true stories about the adventures he had witnessed. But even if his memoirs matched Conan Doyle’s fiction word for word they would not be the same stories… So our world would still not be one where the Holmes stories — the same Holmes stories that Conan Doyle told as fiction — were told as known fact. (pp. 266-267)
So, for him, there might have been a real-life Holmes, but our f-knowledge of him could not have become knowledge had we found out that he was, after all, a real-life character. This seems a mistake to me. I can’t see why the change in status – from f-knowledge to knowledge – couldn’t conceivably take place. I suggest that Lewis is confusing knowledge that one has f-knowledge about Holmes with simply having one’s f-knowledge turn into knowledge upon learning that the Holmes stories, the very same stories, were actually eyewitness reports. That is, contra Lewis, I would hold that your knowing that you have f-knowledge that p is consistent with your piece of f-knowledge turning into knowledge that p — in which case, you cannot, at that later time, still know that you f-know that p. Consider: You’re visiting a nursing home. An old man approaches you and says: “My name is ‘Bond’, ‘James Bond’.” He then proceeds to give you overwhelming evidence that he lived all the 007 adventures up to “Goldfinger”, after which he retired. The stories were published under his nom de plume, “Ian Fleming”. No wonder you thought they were all fiction: He’s lived an ultra-secret life, which is why the stories came out as fiction. The evidence he gives you that he is Bond includes detailed, eye-popping evidence of why the stories never hit the news as fact. If you can so much as imagine that you would be driven into suspension of judgment by his account, isn’t that an intuition that those stories, the very same stories, could conceivably become knowledge for you? You would then, apparently, cease to know that you had f-knowledge of the Bond adventures. It seems wrong to think that you can both know and f-know that p (at the same time). But it certainly remains that you once f-knew what you now know. What’s wrong with that?
Maybe nothing is wrong with that, but it does lead us into our Hard Choice #3: Can storytelling be deceiving with regard to the f-truth? The apparent problem here is that, if you pretend to have made something up while meeting all the necessary objective conditions of an act of storytelling, it would seem that you cannot have failed to make it up. Literary creation seems infallible, no matter what the author may presume he is doing in the act of storytelling! But that would imply that the storyteller cannot be deceitful with regard to the f-truth (even though he can certainly be somehow insincere, that is, he can falsely believe that he is being deceitful). While this seems forced on us by the seemingly unproblematic story about how basic f-knowledge is generated — since it follows from the storyteller’s infallibility with regard to items of basic f-knowledge in the original story — it just seems wrong to think, for instance, that our retired James Bond was not being deceitful in his storytelling (and then coming clean when reporting about the cover-up). (There should be no question that he was insincere, that is, that there was intent to deceive.)
This leads straight into Hard Choice #4: Must storytelling be deceiving with regard to a truth that is represented as an f-truth? This is particularly troubling for anyone who accepts Coincidence. Again, if you don’t, you can resolutely hold, with Lewis, that the real-life Bond did not live that Goldfinger adventure, since only the fictional Bond did, and, appearances notwithstanding, the real-life Goldfinger adventure must not be relevantly related to the fictional Goldfinger adventure, which seems absurd. The real-life Bond was not making anything up when he reported on Goldfinger! But he surely was relying on his storytelling’s being perceived as fictional. He was lying with the truth! If you want to hold that the real-life Bond was somehow deceitful, the only acceptable explanation seems to be that we do tacitly suppose that an f-truth must be false (in the world where the story is told as fiction). That’s why Bond’s original report on Goldfinger seemed safe as an ostensive piece of f-truth. So, Coincidence comes under suspicion.
Hard Choice #5: If you accept Coincidence, why would you want to claim that you can’t both know and f-know that p (at the same time, and assuming p can be known where f-known)? In the Bond case, we were compelled to hold that, once he told you the truth as truth, your f-knowledge turned into knowledge. But, if he was infallible as a storyteller, and if we now have seemingly good reasons (discussed under Hard Choice #3) to regard him as infallible about f-truth, why would you lose your f-knowledge when you gain knowledge? And yet, there is something downright uncomfortable about this, isn’t there?
Lastly, a seemingly unproblematic observation of a general theoretical nature in the epistemology of fiction: F-justified f-true belief may not amount to f-knowledge. You may be f-Gettierized. Suppose you have basic f-knowledge that character C is about to A. When you turn the page, you are led to believe that C has A-ed. You infer that C has A-ed, and your belief is f-true. However, you are reading a defective copy of the novel. When you turned the page f-knowing that C was about to A, you were actually thrown two pages ahead, but had no reason to think that that was the case. We wouldn’t say that you f-know that C has A-ed, even though C has indeed A-ed and you have good, undefeated (but non-entailing) evidence that C has A-ed. Had you believed (or justifiedly believed) that two pages are missing from your copy, you wouldn’t have f-justifiedly believed that C has A-ed. Something surprising might have kept C from A-ing within the missing pages.
Now, back to Virginia and Santa Claus. When Müller posed his question (how can Virginia’s premise-belief have positive epistemic status?) my immediate reaction was: “Her premise-belief cannot have positive epistemic status, because, no matter how reliable her parents may generally be as testifiers, nobody can be reliable about Santa Claus!” We can now see how that initial reaction was a crude expression of a useful point. We can all, of course, be reliable in testifying about fictional characters, provided that our testimony is not a piece of storytelling or otherwise f-justifiedly inferred from what we f-know. We can reliably testify, for instance, that Santa is a beloved fictional character. But we can at best be f-reliable in testifying that he wears a funny red suit.
So, from our foray into the epistemology of fiction, there emerges a fairly simple explanation of the troubling intuitions in the Santa Claus case. It’s certainly not inconceivable that, in claiming that Virginia knows that there will be presents under the tree, the most that we are entitled to claim is that she f-knows that. She certainly can f-know that the presents will be there on Christmas morning on the basis of her f-knowledge that Santa will leave the presents there. Since the language of ordinary knowledge attributions does not discriminate between truths and f-truths, we are inclined to say that she knows what she can only f-know — and we then mistakenly identify her f-true premise-belief with a benign falsehood and agonize over the positive epistemic status that it cannot have acquired (since nobody can be reliable, as opposed to f-reliable, about what can only be f-known).
5. Concluding remarks: Virginia, what’s gotten into you?!
Hard Choice #6: Virginia is obviously very gullible. She has no firm grip on the distinction between fact and fiction. (If she did, it would be obvious to her that there can be nothing even remotely like Santa Claus in our world.) And Mom and Dad are, by hypothesis, generally reliable when testifying to Virginia. Suppose further that Mom and Dad inferentially f-know that Santa is coming on Christmas Eve. It follows from the f-fact that Santa is extremely likely to reward well-behaved children like Virginia. Can Virginia f-know, by Ci-Fi, that Santa is coming? Earlier, it was suggested that only those who can know can also f-know, and Virginia certainly cannot have justified (as opposed to f-justified) beliefs about Santa. But, unless something very much like Ci-Fi is defensible, we haven’t come anywhere near explaining how Virginia can f-know that there will be presents under the tree on Christmas morning. (Notice, she certainly cannot be f-justified in her premise-belief by reasoning inductively about her parents’ f-reliability, since, by hypothesis, she has no concept of f-reliability.) And that may seem to be the closest we can get to explaining why we are — as we undoubtedly are — inclined to say that she knows that there will be presents on Christmas morning.
Hard Choice #7: There is an alternative account of the temptation to say that Virginia knows that there will be presents under the tree that does not involve a confusion between knowledge and f-knowledge. It may plausibly be held that Virginia has gathered much evidence of her parents’ reliability as testifiers. On very many occasions, what they said turned out to be true. Rarely did what they asserted prove false. So, she has much evidence available to her for a good inductive argument in support of her premise-belief (that Santa is coming). Isn’t that enough for her to be justified in her premise-belief? The problem with this is twofold. First, it is far from obvious, given the degree of gullibility required for a belief in Santa, that we can reasonably expect a child like Virginia to avail herself of the evidence and use it effectively in support of her premise-belief (on what would then be a reductionist account of her testimonial justification). So, it remains more reasonable to conclude that Virginia could not be doxastically justified in holding her premise-belief.  Moreover, we could simply stipulate that she does not avail herself of the evidence, and it wouldn’t be an unreasonable stipulation. Thus, a reductionist account according to which she is justified in holding her premise-belief would have to resort to the unpopular view that the mere availability of the evidence is enough for her to hold a justified belief. Second, even if we let Virginia use the evidence of her parents’ reliability, thus giving our understanding of the case a reductionist twist, Virginia must still remain gullible enough to believe in Santa – which implies that she must remain insensitive to the overrriding effect of the (overwhelming) evidence of Santa’s non-existence on her prima facie justification from the inductive argument. So, on a reductionist account of testimonial justification, Virginia must hold an unjustified premise-belief. But, hold the thought that we are not going anywhere! It may be claimed – not implausibly – that it takes quite a bit of epistemological reasoning to conclude that Virginia cannot be justified in her premise-belief. Maybe what accounts for the commonsense reaction to the case – according to which Virginia does not unreasonably hold her premise-belief – is the fact that the non-philosophical folk (all of us in our non-philosophical moments) tend to attribute to her the kind of inductive reasoning that mature adults would deploy in support of her premise-belief. What then goes wrong – very wrong – is that the non-philosophical folk miss the part about epistemic defeat and cut Virginia the kind of slack that epistemological theory would deny us all. This may indeed explain the intuitive appeal of the claim that Virginia’s premise is a benign falsehood. Trouble is, on this account of the commonsense reaction, Virginia must still remain unjustified in her premise-belief. So, we haven’t really made any progress in explaining how Virginia can know that there will be presents under the tree. Must we throw in the towel as regards her knowing anything and rest content with this attractive explanation of why people tend falsely to think that she does know?
Being one who finds it hard to shake the conviction that Virginia knows that there will be presents on Christmas morning – regardless of whether Ci-Fi is tenable and she also f-knows – I’m as puzzled about the Santa Claus case as anybody can be. 
1. See Klein’s 2007. The main features of this forthcoming paper were disclosed by him in 2004 and discussed at Certain Doubts in Jon Kvanvig’s posts “Klein on Useful Falsehoods”, “Klein on Useful Falsehoods II” and “The Significance of Useful Falsehoods”, and in my post “On Useful Falsehoods”. Fritz Warfield joined the discussion with his 2005 paper.
2. I used the example in my unpublished 2003 talk at the Central States Philosophical Association meeting in Chicago, where EJ Coffman was my (insightful) commentator.
3. Pritchard goes on to discuss two other qualified versions of Credulism — Modest Credulism and Modest Restricted Credulism. But the distinctions shouldn’t make any important difference in what follows, though they certainly matter for a more extensive discussion in the epistemology of testimony. Bare Credulism is implied by a claim (derived from Tyler Burge’s work and opposed by Elizabeth Fricker) that Matthew Weiner (2003) has recently argued for: “We are justified in accepting anything that we are told unless there is positive evidence against doing so.”
4. I certainly don’t mean to imply that Credulism will ultimately prove untenable. I simply acknowledge the fact that it’s taken quite a bit of work to turn it into a view worth discussing. It is, to my knowledge, universally conceded that externalism is on the defensive in the epistemology of testimony. This is consistent with believing, as I do, that a much qualified version of Credulism is prevalent in the contemporary debate, since it has been made compatible with Humean reductionism (specifically, some form of “local reductionism”). The main argument for Credulism is briefly discussed below.
5. Some writers seem to simply assume that the correct phenomenology of testimony acceptance in those kinds of situations calls for thinking that one normally believes what one is told in the absence of grounds for distrust. See Weiner 2003 for an instance. For another, consider this passage in James B. Freeman’s 2005 work on the issue:
Just as in perception, when we are appeared to in a certain way through the senses, we automatically form certain beliefs as a consequence of our constitution or design plan, so when we are appeared to “testimonially” we are constitutionally set to assent to the belief formed through this experience… As a presumption for the reliability of perception is a first principle of common sense, so a presumption for the reliability of relying on testimony is also a first principle of common sense. As such, there is no need to argue for these principles and indeed such an argument would be wrongheaded. (292-3)
The problem is that this is woefully at odds with our institutional practices, which constantly and essentially call for the scrutiny of credentials, and, if anything resembling Freeman’s description is even ballpark-right, it must be a mystery what sets us on our distrustful ways so early in life. Contrast that descriptive credulist assumption with the discussion in Jonathan Adler 2002, pp. 143-144:
By and large, hearers accept speakers’ testimony, unless hearers have specific reason to object… When the stranger tells me, in response to my inquiry, to take the number 4 downtown to 14th Street…I, as the hearer, will generally do so because I accept his word. Admittedly, the uniformity that we observe is imperfect, and the behavioral data somewhat equivocal… The behavioral data are equivocal because when the hearer conforms to the default rule, he may not truly accept the speaker’s word. If the hearer is under time pressures, he may have no practical choice but to do what the speaker says, and a hearer can appear to be accepting the speaker’s word, but not do so in the privacy of his mind…
In spite of this, Adler goes on to claim that,
normatively and statistically, speakers expect hearers to take their word. If I am offered directions and do not challenge them, then if the speaker observes me not following his directions, he will be offended (“Why didn’t you trust me?”); and if he confronts me, I am shamed.
I haven’t been able to confirm that the adverb “normatively” is correctly used here. On the contrary, the classroom experiment indicates that it doesn’t take much coaxing at all to get people to admit that it is unreasonable for a stranger to expect to be trusted in the absence of evidence of reliability. So, the statistical data seems of little value here. It doesn’t take much discussion for people to reconsider their initial reaction to skeptical inquirers. Moreover, insincerity is unpopular — occasionally perceived as immoral — even when socially valuable. Also, the non-philosophical folk tend to confuse epistemic with prudential justification, but seem to have no difficulty embracing the distinction when confronted with it. It shouldn’t be a surprise that it takes the controlled environment of the philosophy class to make our subjects reconsider their instinctive behavior.
6. For good criticism of the asymmetry in Fricker’s reductionist stance, see Weiner 2003.
7. Another attractive alternative will be mentioned shortly (note 8 below).
8. Stephen Hetherington’s “epistemic gradualism” is a very attractive alternative to my proposal here. (See Hetherington 2001 and 2005.) Exploiting the well-known fact that justification is a graded phenomenon and that we are notoriously incapable of telling, non-arbitrarily, where the cut-off point is between justification that is insufficient for knowledge and that which is, he puts forward a gradualist view of knowledge with mere true belief as its lower bound:
We may accept…that no step along the spectrum is a cut-off point between knowing and not knowing…[A]ny true belief is already knowledge. Justification makes it better knowledge — and better justification makes the belief even better knowledge still. (2001, 148-9)
On that basis, he is engaged in our discussion about how best to account for epistemic growth:
When a young child is told that the earth orbits the sun, she thereby comes to know that it does. If this is not knowledge, her education at that stage is not the imparting of knowledge. But of course that is what it is…Other things being equal, being told by five sources — rather than just one — that the earth orbits the sun improves one’s knowledge of its doing so. Don’t you know better now that the earth orbits the sun than you did as a child? It is obvious that you do. (2001, 148-9)
I can’t discuss Hetherington’s gradualism at length here and would hate to oversimplify it. It is an attractive alternative. But I can’t say that I’m fully persuaded that he has the best overall account of the intuitions in play when we use the language of knowledge in describing that crucial transition in our epistemic growth. I have preferred to account for it as the crossing of a threshold: At some point, the child is equipped with whatever it takes for us to describe some of her true beliefs as knowledge. (This will still be consistent with speaking of the improvement of one’s knowledge, as per Hetherington’s admonition.) The crucial element is the desired reliability check, whatever the best explanation of that may be (the acquisiton of beliefs about the reliability of sources, or just some ability to detect unreliability, whatever). Granted, Hetherington’s view will make the most of out this admission of ignorance. (It’s a good thing that it will; it’s what gives us a viable alternative.) On the other hand, his view may suffer when confronted with the time-honored, intuitively well-founded and apparently sharp distinction between knowledge and true belief. I wonder if the good premise concerning the undoubtedly graded phenomenon, epistemic justification, will safely deliver Hetherington’s conclusion about the existence of a degree of knowledge that coincides with merely true belief. I suspect that the linguistic evidence from ordinary parlance in cahoots with the argument from vagueness is no firm ground for the conclusion about “minimal knowledge” (true belief). And I must leave it at that.
9. From the standpoint of a plausible phenomenology of belief, I can’t see that there is any good reason to assume that belief in what is ostensibly an f-truth is a propositional attitude to be distinguished from ordinary belief (in truths which obtain in the world of the believer). I think Lewis would agree with this. But his acceptance of a notion of “make-believe learning” (in his 1983), derived from Kendall Walton’s work on “make-believe truth”, may indicate otherwise. In the case of Gareth Evans’ work on the issue (1982, ch. 10), Walton’s influence does lead to the positing of “make-believe propositional attitudes”. I shall not discuss Evans’ view of “make-believe belief” here. I shall only note, in passing, that it strikes me as ill-founded. (See note 14 below.)
10. Here, I will only go so far as to offer you the main features of a foundationalist epistemology of fiction.
11. As we shall shortly see, the storyteller may not be able to cause any basic f-justified TBB’s. The fact that some TBB’s can’t be f-justified even if they are appropriately received from the storyteller — because their f-justification is defeated by one’s f-justification for some inferential beliefs — suggests that this is a form of moderate foundationalism (with both infallible and defeasible f-justification generated by the author at different times). More on this shortly.
12. For the remainder of our discussion, I will rest content with focusing on this very rough formulation of Ci-Fi. The tough questions ahead wouldn’t significantly be affected by more sophisticated formulations of that claim.
13. The question of whether our prototheory provides for any difference between the storyteller and the madman will shortly be tackled.
14. If you are inclined to think that it is a fundamental mistake to think of storytelling as the offer of a kind of testimony (testimony purporting to convey f-truths), consider again the cost of buying your way out of here: Very clearly, storytelling seems to cause beliefs, and those beliefs seem very ordinary indeed. (Notice that, for us, what one may want to call an “f-belief” is an ordinary belief that is a case of f-knowledge. There are no special propositional attitudes, which is why I’ve avoided the term “f-belief”.) A fortiori, the other relevant propositional attitudes will also be a part of the picture. We f-justifiedly disbelieve that Don Quixote used to ride a motorcycle, and we f-justifiedly suspend judgment on whether he was once cared for by a Chinese nanny. But, again (see note 9 above), one will find an alternative in Evans (p. 359):
A story-teller pretends to tell (inform) us about things. He pretends to be informed about these things. (That is, he pretends to be informationally related in some way or other to the events he relates; although the precise relation need not enter into the pretence.) We, hearing him, are prone to carry on the pretence…we pretend to have been told of these things (to know them by testimony).
I think this will yield a counterintuitive explanation of the data. I’m not pursuing the comparison at length here. Still, consider: Why do we need a notion of make-believe truth if the storyteller is only pretending to be informative? Why not simply have the truth (falsehood) as the object of pretence? Also, if pretending to believe (“make-believe belief”) is incompatible with full-blown belief, as Evans seems to suppose, shouldn’t it be under your voluntary control? Is the propositional attitude that you have toward the proposition that Quixote battled windmills under your control? Can you just pretend to believe that he did? Mere pretence surely is under voluntary control.
15. Lewis tacitly recognized the social aspect of f-knowledge as a part of his rationale for letting the set of “collective belief worlds” of the community where the story is told as fiction (defined by the beliefs that are “overt” in that community) constrain his analysis of f-truth.
16. There is, of course, a kind of spoof that exploits the outrageous misplacement of well-known fictional characters. If the misplacement could yield f-knowledge, it wouldn’t be funny.
17. See Audi 2003, ch. 7, for a sketch of a version of moderate foundationalism.
18. The classic presentation of the doxastic/propositional justification distinction is in Roderick Firth 1978.
19. A special thank-you goes to my colleague Felipe Müller for the question that set these cogitations in motion. For the insightful comments they so generously offered, I thank Rodrigo Borges, Stephen Hetherington, Jennifer Lackey, and Ed Ribeiro.
Adler, Jonathan E. 2002. Belief’s Own Ethics (MIT Press).
Audi, Robert. 2003. Epistemology, 2nd edition (Routledge).
De Almeida, Claudio. 2004. On Useful Falsehoods. Certain Doubts,
___. 2003. Benign Falsehoods: A Lesson about Inferential Knowledge. Unpublished talk at the Central States Philosophical Association, Chicago.
Evans, Gareth. 1982. The Varieties of Reference (Oxford University Press).
Firth, Roderick. 1978. Are Epistemic Concepts Reducible to Ethical Concepts?. In A.I. Goldman and J. Kim, eds., Values and Morals (Reidel).
Freeman, James B. 2005. Acceptable Premises: An Epistemic Approach to an Informal Logic Problem (Cambridge University Press).
Fricker, Elizabeth. 2004. Testimony: Knowing through Being Told. In Ilkka Niniluoto, Matti Sintonen and Jan Wolenski, eds., Handbook of Epistemology (Kluwer Academic Publishing).
Goldman, Alvin. 1999. Internalism Exposed, The Journal of Philosophy 96.
Hetherington, Stephen. 2005. Knowing (How It Is) That P: Degrees and Qualities of Knowledge. In Claudio de Almeida, ed., Perspectives in Contemporary Epistemology, Veritas, vol. 50, No. 4 (Porto Alegre, Brazil), downloadable for free at http://revistaseletronicas.pucrs.br/veritas/ojs/viewissue.php?id=4&locale=en .
___. 2001. Good Knowledge, Bad Knowledge (Oxford UP).
Klein, Peter. 2007. Useful False Beliefs. In Quentin Smith, ed., New Essays in Epistemology (Oxford UP, forthcoming).
Kvanvig, Jonathan. 2004. Klein on Useful Falsehoods. Certain Doubts, http://fleetwood.baylor.edu/certain_doubts/?p=161#more-161 .
___. 2004a. Klein on Useful Falsehoods II. Certain Doubts, http://fleetwood.baylor.edu/certain_doubts/?p=162#more-162 .
___. 2004b. The Significance of Useful Falsehoods. Certain Doubts, http://fleetwood.baylor.edu/certain_doubts/?p=164#more-164 .
Lackey, Jennifer. 2007. Learning from Words (Oxford UP).
Lewis, David. 1983. Postscripts to “Truth in Fiction”. In his Philosophical Papers, Volume I (Oxford UP).
___. 1978. Truth in Fiction. In his Philosophical Papers, Volume I (Oxford UP).
Pritchard, Duncan. 2004. The Epistemology of Testimony. In Ernest Sosa and Enrique Villanueva, eds., Philosophical Issues, 14, Epistemology, 2004 (Blackwell Publishing).
Warfield, Ted. 2005. Knowledge from Falsehood. In John Hawthorne, ed., Philosophical Perspectives, 19, Epistemology, 2005 (Blackwell Publishing).
Weiner, Matthew. Accepting Testimony, The Philosophical Quarterly, vol. 53, No. 211.