Evidentialism and the problem of stored beliefs

Philos Stud (2009) 145:311 324 DOI 10.1007/s11098-008-9233-1 Evidentialism and the problem of stored beliefs Tommaso Piazza Published online: 10 May 2008 Ó Springer Science+Business Media B.V. 2008 Abstract Many stored beliefs, like beliefs in one s personal data or beliefs in one s area of expertise, intuitively amount to knowledge, and so are justified. This uncontroversial datum arguably tells against evidentialism, the position according to which a belief is justified if it fits the available evidence: stored beliefs are normally not sustained by one s available evidence. Conee and Feldman have tried to meet this potential objection by relaxing the notion of available evidence. According to their proposal, stored beliefs are dispositionally justified, because they are justified by the evidence one has the disposition to retrieve; such evidence, as a consequence, is to be characterize as available, though in a derivative sense. Goldman has criticized this proposal, by offering a counterexample to the claim that a disposition to generate a piece of evidence may qualify as a justifier. In this paper I critically examine two possible replies to Goldman s example stemming from Conee and Feldman, and finally propose my own, based on a distinction, inspired by Audi, between dispositional evidence and the disposition to have evidence. Though this proposal differs from Conee and Feldman s one, I will conclude that it fits pretty well their intuitions. Keywords Internalism/Externalism Evidentialism Dispositional evidence Stored beliefs Evidentialism is the theory of justification according to which the justification of a person s belief at t supervenes on the evidence that person has at t. Evidentialism offers a very natural view on justification. It asserts that the justificational status of a person s beliefs depends on the quality and the quantity of evidence at her disposal. The view is made plausible by examples. Such examples normally show two T. Piazza (&) Department of Philosophy, University of Porto, Via Panoramica, Porto, Portugal e-mail: tpiazza@letras.up.pt

312 T. Piazza different subjects entertaining the same belief with a different degree of justification, and suggest that the difference can be explained by considering the different quality and quantity of information available to both 1 (Conee and Feldman 1985, 2001, 2004). Its initial plausibility notwithstanding, this view has received in recent time several criticisms. In this paper I want to examine in details one of these criticisms, recently advanced in Goldman (2002), and try to implement an answer offered by Conee and Feldman I find insufficient. Among the things we know, so reads a very plausible suggestion, there are things that we have stored in our memory. For instance, I do not need to think explicitly to the PIN code of my mobile telephone to be credited knowledge that it is #### (believe me, I know my PIN code, it is just for prudential reasons that I don t tell you which one is it). Knowledge of it can be credited also if all the evidence I happen to consider (henceforth, O-evidence) is entirely devoted to a different topic, as it was the case before I decided to present this example. However, if it is so, a plausible difficulty for the evidentialist is that some minutes before I had no evidence supporting the belief that my PIN code is ####. So, the evidentialist is apparently committed to denying that I knew it (see Goldman 1999, p. 278 for a clear statement of this criticism). Reliabilism, the rival externalist view according to which, roughly, a belief is justified only if it is the outcome of a reliable belief forming process clearly does not have this problem: I can be credited knowledge of my PIN code just provided that the way I acquired the belief is reliable enough. So, it would seem, the case of stored beliefs intuitively amounting to knowledge allows the reliabilist to score one point against the evidentialist. *** The aim of reconciling evidentialism with the idea that many stored beliefs are known and justifiably entertained is not new. Feldman, for instance, has anticipated Goldman s objection and has denied that the justificational status of stored beliefs cannot be accounted for in evidentialist terms (Feldman 1988). The plausible line of response he pursues is to relax the conditions under which something is to count as a piece of available evidence. In particular, Feldman has proposed to distinguish among occurrent and dispositional senses of epistemic terms, and to account for the epistemological status of stored beliefs in dispositional terms. Though Feldman just distinguishes between occurrent and dispositional senses of knowledge, it is sensible to read him as entailing that such distinction should be accepted also for the related notion of justification. Feldman says that: We can, however, introduce a dispositional sense of know in which such things [propositions like that Washington is the capital of USA] are known. That sense might be roughly characterized in terms of the occurrent sense: a person knows a thing dispositionally provided the person would know it occurrently if he thought of it. Since the thought that Washington is the capital 1 Conee and Feldman (2001) propose the following principle to specify the internalist approach to justification: «M. If any two possible individuals are mentally alike, then they are justificationally alike, e.g., the same beliefs are justified for them to the same extent» (2001, p. 235).

Evidentialism and the problem of stored beliefs 313 would, presumably, be accompanied by an awareness of justifying evidence, this fact can be known dispositionally by most of us (Feldman 1988, pp. 98 99) Since knowledge entails justification, it seems that also dispositional knowledge (D-knowledge) will entail dispositional justification (D-justification). How are these notions to be understood? The quotation seems to suggest that, on the one hand, S s belief that p at t is occurrently justified (O-justified) if it fits S s occurrent evidence (O-evidence) at t. On the other hand, S s belief that p is dispositionally justified (D-justified) at t if it doesn t fit S s O-evidence at t, but S would have O-evidence of the appropriate kind (this derivative evidence amounting to D-evidence) were she to think of p (see Feldman 1988, pp. 98 99). In the example above, I was too busy with the paper to entertain any evidence that could support the belief that the PIN code is ####. So my O-evidence didn t encompass anything supporting such belief. However, I had the disposition, upon considering my PIN code, to generate O-evidence of the appropriate kind: in the sense just presented, I had D-evidence. Since the quotation also seems to suggest that a true (stored) belief, if D-justified, may count as D-knowledge, it shows that the evidentialist is not banned from acknowledging that I did know my PIN code, even if this knowledge has a somehow derivative status, that of D-knowledge. This interpretation of Feldman s reply raises two important concerns, respectively related to the notion of dispositional justification and the notion of dispositional knowledge. Let us begin by taking into account Goldman s worry, according to which no clear sense can be attached to the suggestion that a belief may be D-justified in the sense just adumbrated. The discussion of what must be added to D-justification in order to turn a true belief into D-knowledge shall not occupy us until the final part of the paper. A. Goldman has called into question Feldman s answer (Goldman 1999, pp. 278 279, 2002, p. 9). He has written: if having a disposition to generate conscious evidential states qualifies as a justifier of a belief, why wouldn t this extend from memorial to perceptual dispositions?. Before considering in further details his counterexample, let us state briefly Goldman s interpretation of Feldman s proposal. Though Goldman does not explicitly recognize it, the preceding quote arguably suggests that Goldman understands Feldman s notion of dispositional justification in the following way: (D-j) If O-evidence E would O-justify the belief that q, the disposition to generate E D-justifies the belief that q; Now, consider the following example. Tom is traveling by train, and gets asleep for a while. Unbeknownst to him he is still sleeping his favorite soprano, Monica Bacelli, takes seat in front of him. Suddenly, a train announcement awakes Tom. Tom is still sleepy, and for 10 s he does not open his eyes. So, for 10 s, he is conscious, yet the evidence at his disposal does not encompass visual information concerning his environment. In particular, he does not enjoy perceptions that would inform him of the presence of the Italian soprano. What is the justificatory status of the belief that there is Monica Bacelli sitting in front of him during these 10 s?

314 T. Piazza Unquestionably, Tom s situation is such that he has the disposition, upon opening his eyes, to enlarge his evidence so as to make the belief that Monica Bacelli is sitting in front of him justified. More than this, such evidence would O-justify this belief. Acceptance of (D-j), then, commits one to grant to Tom D-justification. Nonetheless, Goldman correctly observes, there seems to be no available sense in which Tom is justified in entertaining the belief. So, he concludes, (D-j) must be rejected. An easy way to elude Goldman s counterexample, it might be suggested, is simply to deny that Conee and Feldman are committed to accepting (D-j). The argument would read as follows: according to the counterfactual definition presented in Feldman 1988, something qualifies as a dispositional justifier for S s belief that p only provided that, were S to consider whether p, she would generate O-evidence that the belief that p fits. Arguably, under this definition a disposition to perceive one s environment doesn t qualify as a D-justifier: were Tom to consider the proposition that Monica Bacelli is sitting in front of him, while his eyes are still closed, he wouldn t generate the relevant (perceptual) evidence in the light of which his acceptation of the proposition would be O-justified. In order to do so, Tom should have to open his eyes, and acquire the perceptual evidence for the first time. On the other hand, were I to think of my PIN code, my belief that it is #### would be probably accompanied by O-evidence of the appropriate kind, arguably consciously entertained memories of the fact that whenever I type #### on my mobile phone, it switches on correctly. So, though (D-k) goes unchallenged, Conee and Feldman s position arguably does involve the commitment to the following different definition of D-justification. (D-j*) Evidence E qualifies as a D-justifier for S s belief that p iff, were S to consider whether p, S would generate O-evidence E that the belief that p fits 2. Though with (D-j*) in the place of (D-j) Conee and Feldman s position is not anymore open to Goldman s counterexample, (D-j*) is not immune from different kinds of problems. The gravest shortcoming, in particular, seems to be that (D-j*) may characterize as a D-justifier evidence which, in no relevant sense of the term, a subject may be told to possess. Just consider the following case: Alex is a clever maitre with very good memory and mathematical skills; one day, as usual, he stands at the main entrance of the restaurant. Since every person entering or leaving the restaurant must have done it in front of his eyes, Alex s intellectual abilities predispose him, at every given time, to calculate the number of the persons who are inside at that moment. So we can assume that Alex, if asked how many people are inside the restaurant, would perform a calculation whose result would occurrently justify him in delivering a given answer. So, if (D-j*) is correct Alex is justified even before performing the calculation. However, it seem intuitive that he is not. Evidence one does not yet possess as the evidence one would generate were one to 2 I am indebted to an anonymous referee for the suggestion that Conee and Feldman could elude Goldman s counterexample by using (D-j*) to characterize D-justification.

Evidentialism and the problem of stored beliefs 315 perform a calculation cannot be evidence in light of which one is (in some sense) already justified 3. A better answer, it might then be suggested, can be evinced by a recent reply proposed by Conee and Feldman to Goldman s argument. Here is the relevant passage from Internalism Defended: The idea behind the current proposal is not what Goldman criticizes here. It is not that any conscious mental state that one is disposed to be in counts as evidence. The idea is that some non-occurrent states that one is already in, such as non-occurrent memories of perceptual experiences, are stored evidence. Presently having this stored evidence justifies dispositionally some non-occurrent beliefs that one already has. The train passenger does not have the evidence that he would receive were he to open his eyes. The dispositional state that he is in, his disposition to see the landscape by opening his eyes, is not stored evidence for propositions about the landscape. It is a potential to acquire evidence, and that is crucially different. (Conee and Feldman 2001, p. 244) Conee and Feldman s characterization of dispositional justification does not anymore mention the counterfactual analysis displayed by (D-j*). To identify what determines dispositional justification the authors just refer to stored evidence, which they suggest is exemplified by non-occurrent memories of perceptual experiences. So, they might be read as suggesting that the notion of dispositional justification is encoded by the following principle: (D-j**) Evidence E qualifies as a D-justifier for S s belief that p iff E is stored evidence, and S has a disposition to retrieve E and so to generate O-evidence that the belief that p fits. Basically, the suggestion conveyed by (D-j**) is to restrict the dispositional states mentioned in (D-j) and counterfactually defined in (D-j*) to the dispositions to retrieve stored evidence. Once such restriction is in place, both counterexamples do not matter any more. On the one hand, Goldman s counterexample draws on a disposition to generate a perceptual state, not on a disposition to retrieve stored evidence. On the other hand, the second counterexample draws on evidence 3 Although he doesn t seem to worry much about it, Feldman arguably recognizes the problem. In Feldman 1988 we read that there may be some things that I do not know now, but if I were to think of them I would then come to believe them, and come to have evidence for them, for the first time (p. 99). Feldman is here suggesting that the counterfactual characterization of epistemic terms presented above may run into troubles whenever one s epistemic situation disposes one to form for the first time a belief in a proposition, and to entertain for the first time evidence which supports such belief. For in that situation it is true that the subject would know [that proposition] occurrently if he thought of it, yet in no available sense it is true of the subject that she does already know it. As I read this passage, Feldman explains the latter intuition by noticing that in the cases under discussion a person would fail two meet two necessary conditions for knowledge, because she wouldn t believe a proposition, and she would fail to possess evidence bearing on that proposition. So, a fortiori, Feldman is arguably committed to denying that evidence one has not yet considered for the first time, as the deliverance of Alex s possible calculation, is evidence in light of which one is in some sense justified.

316 T. Piazza provided by a calculation process, which doesn t count as stored evidence in any sense of the term. I think that (D-j**) correctly points in the right direction: the evidential import of stored evidence is clearly asymmetric to the evidential import of the evidence one is just in a good position to acquire. The latter kind of evidence is not evidence one has, while old evidence, so long as it has been stored and retained, is still evidence one possesses. However, I think that Conee and Feldman s identification of dispositional evidence with stored evidence must be supplemented with additional arguments if it is to ground a convincing reply to Goldman s challenge. The reason is as follows. The original problem, emphasized by Goldman, concerns the justificational status of stored beliefs that, on an intuitive basis, are to be regarded as amounting to knowledge. As we have already seen at the beginning of the paper, the evidentialist lets the justificational status of a belief depend on the quality and quantity of available evidence it fits. From the evidentialist perspective, then, the problem of the justificatory status of stored beliefs intuitively amounting to knowledge can be dealt with only by showing that the kind of things in the light of which these beliefs are justified count, even if in a derivative sense, as available evidence. So, the challenge the evidentialist has to face is to explain in what sense, if any, the stored evidence in the light of which stored beliefs are typically justified counts as available evidence. Goldman s objection apparently rules out as a possible answer the observation that such evidence is made available by the disposition to retrieving it: being the object of a disposition to make the evidence occurrent, and available in more central sense, is not sufficient to discriminate among the evidence to which it is sensible to attribute an evidential import from the one to which it is sensible not to attribute it. So, Goldman s argument leaves the evidentialist with the task of explaining in a better way what it is, of a body of stored evidence, that makes it available and evidentially relevant. By itself, (D-j**) doesn t discharge this explanatory role: it just conveys the claim that the kind of things in the light of which stored beliefs can be justified count as available evidence. Yet, unless the restriction to stored evidence that it encapsulates is motivated in the light of some criterion, whose satisfaction somehow explains why a disposition to retrieve stored evidence makes this evidence available, while any other disposition does not, this is a trivial result. In the absence of such criterion the proposal of (D-j**) can simply be dismissed as introducing an ad hoc distinction that has no explanatory value. In fact this is exactly what Goldman does, when he anticipates Conee and Feldman s proposal, and queries its explanatory value in the following terms: Feldman might reply that there is an important distinction between memorial and perceptual dispositions; but it is unclear on what basis a principled distinction can be drawn (Goldman 2002, p. 22, fn. 10). It is my opinion that, much in the direction hinted at by Conee and Feldman, there is a principled distinction to be drawn. The simple answer I shall argue for is that a body of evidence that is not occurrently entertained nonetheless constitutes available evidence just provided that its possession can be credited dispositionally. Firstly, I shall explain the notion of possessing a body of evidence dispositionally by generalizing R. Audi s distinction between a dispositional belief (which I shall treat

Evidentialism and the problem of stored beliefs 317 as constituting a piece of doxastic evidence one possesses dispositionally) and a disposition to form a belief. While having a disposition to form a belief (the disposition to acquire evidence) is not a way of possessing evidence, it is arguable that believing dispositionally (having evidence dispositionally) is a way of possessing it. Secondly, I shall argue that a body of stored (doxastic and nondoxastic) evidence, granted the satisfaction of certain conditions of psychological accessibility, constitutes evidence whose possession can be credited in the dispositional sense. Thirdly, I shall point out that the notion of dispositional evidence (evidence that is possessed dispositionally) is immune from Goldman-like counterexamples, which draw on epistemic scenarios where a subject has just a disposition to acquire evidence, and does not possess evidence dispositionally. *** Let us start by clarifying the needed vocabulary. Coming to entertain an O-belief may be the result of a process of belief formation. I may come occurrently to believe that it is raining because, upon being addressed a question as to whether it is raining, I simply look out of the window, see that it is raining, and appropriately form the belief that it is raining on the basis of that evidence. However, suppose that matters were slightly different. I already acquired the relevant evidence, because, while drinking my morning coffee, I was absentmindedly looking through the window while thinking of some pressing question. Since it was raining we can suppose that I registered this bit of visual information, while abstaining from forming the corresponding belief 4. I was simply mentally too busy to do so. After a while, at t, my girlfriend asks me whether it is raining. This time I do not need to look outside the window. The question calls back the visual information I registered some instant before and spontaneously return the answer that it is raining. Also this time my O-believing at t is the result of a process of belief formation. The difference with the first case is that I already had at my disposal information that disposed me to form this belief. Upon consciously reflecting on such information, and considering whether it is raining, I had the disposition to form the belief that it is raining. In this case we will say that my forming the corresponding O-belief is causally explained by the realization of my disposition to believe that it is raining (B-disposition). B-dispositions, according to R. Audi, should be sharply distinguished from dispositional beliefs (D-beliefs). Consider the following, slightly modified scenario. It is always me, with my coffee, sadly looking at the rain falling out of the window. I am to be at the University in 1 h, the bus is on strike, and the only alternative to catching a cub is to ride my bicycle. I process all the information at my disposal and decide to ride the bicycle. Perhaps, I say to my self, the rain will stop. After a while I start reflecting on the reason why I am to be at the university so early. I have to deliver a talk. Accordingly, I start mentally to check whether everything is in order: I remember the main steps in my paper, I have the power-point file stuffed in my USB stick, etc. For the 4 It is reasonable to suppose that we actually form just a very small part of the beliefs that are justified in the light of the informational states we are in. As Audi has remarked, «there seems to be a natural economy of nature perhaps explainable on an evolutionary basis that prevents our minds being cluttered with the innumerable beliefs we would have if we formed one for each fact we can see to be the case» (Audi 1998, p. 22).

318 T. Piazza time being, the belief that it is raining is not anymore occurrent. It becomes a stored belief 5. Accordingly, when my girlfriend asks me whether it is raining, and I spontaneously return the answer that it is raining, I am not forming the belief that it is raining; rather, I am just activating a D-belief of mine. To explain the difference Audi makes use of a computer analogy: O-beliefs are like things displayed on a computer screen. D-beliefs are analogous to things that are not visible in the computer screen, yet stored in the computer s memory. B- dispositions are like things a computer screen would display, were the computer to make certain calculations. In the former case, D-beliefs are already there: «[they] need only be brought to the screen by scrolling a simple retrieval process in order to be used» (Audi 1994, p. 420). In the latter case, a belief one has a disposition toward in not already there: «the raw materials, which often include inferential principles, are present, but the proposition is not yet in the memory bank or on the screen» (ibidem). Importantly, Audi s distinction seems to generalize also to non-doxastic kinds of evidence. Firstly, if subject S acquires visual information I concerning her environment, we might suppose that she has I dispositionally at her disposal also when she doesn t consciously entertain it (in the example above, I had dispositionally at my disposal the visual information I acquired looking through the window, even if I was consciously thinking just of my talk). Secondly, if subject S has not yet acquired it, but her situation disposes her to acquire such information (as in the case of Tom before he opens his eyes), we will say that she has the disposition to acquire such information. In what follows, we will generally talk of possessing evidence dispositionally (possessing D-evidence) and of being disposed to acquire evidence (being E-disposed). The distinction between D-evidence and E-dispositions seems to be very relevant for assessing (D-j**). D-evidence and E-dispositions may be confused because their manifestations are the same: to take just the example of doxastic evidence, both D-beliefs and B-dispositions may both result in a subject spontaneously coming to entertain an O-belief under the same circumstances. Their status is however completely different. A D-belief is a belief a subject already entertains, though in a derivative way, while a B-disposition is just a disposition to form, when the circumstances require it, a new belief. In the same way, a piece of D-evidence is evidence one already possesses, while having a disposition to gather evidence, somehow grounded in one s epistemic situation, is not to have evidence. From an evidentialist point of view, this difference will naturally lead to assess the evidential import of both kinds of state differently. A D-belief is available in a sense in which a B-disposition is not. 5 It is questionable that belief storage is always the storage of an O-belief as in the case at issue. If you distinguish between simple seeing that kind of seeing reported by Sam sees the tree in the garden and propositional seeing that kind of seeing reported by Sam sees that there is a tree in the garden, you re likely to think that the latter kind of seeing involves forming the belief in the proposition specified by the that-clause in the statement attributing to Sam the perceptual state. In this case a belief has been formed, but it is questionable that it needs to be occurently entertained. R. Audi, for one, suggests that the belief can be directly turned into stored belief and, therefore, that it needs not be O-belief. See Audi (1994, pp. 420 421).

Evidentialism and the problem of stored beliefs 319 What is believed is, unless temporarily forgotten, presuppositionally available in thought and discourse, e.g. as a premise for inference or as spontaneously assertible in topics on which the proposition in question bears. What one is merely disposed to believe is only indirectly available, say through considering the proposition or placing it in the light of one s background beliefs. Some kind of realizer is required to produce belief formation (Audi 1994, p. 424). Difference in availability between D-evidence and E-dispositions actually explains difference in evidential import. On the one hand, I cannot be said to be justified in the light of evidence I do not have, even if my epistemic situation is such that, were certain realizing conditions to be fulfilled, I would acquire for the first time that evidence. Actual acquisition of such evidence seems to be required for it to count as a justifier. On the other hand, my having D-evidence for a proposition explains why I am justified in believing whatever the latter evidence sustains. In the light of the foregoing considerations, it should be clear why Conee and Feldman s restriction to stored evidence is entirely appropriate. The reason why a piece of stored evidence has epistemic significance is that, arguably, storage is a way of possessing evidence dispositionally, and possessing evidence dispositionally is a way of possessing evidence. The basic suggestion here is that what we currently mean when we say of some S that she remembers that p, or that she has some nondoxastic memory of a certain situation, is nothing but that she once acquired the belief that p or that she once acquired some information concerning a situation, and that the belief, or the non-doxastic information is retained and available in the sense that S, to say in C. Ginet s words, would manifest it given suitable motive to do so (Ginet 1975, p. 145). This might encourage the further thought that S has stored evidence E iff S has D-evidence E; in other words, it might encourage the thought that the notion of stored evidence and the notion of dispositional evidence are at least extensionally equivalent. I do not think that stored evidence and dispositional evidence are extensionally coincident. Arguably, in fact, actual storage just provides a necessary condition for dispositional availability, not also a sufficient one. As a matter of fact, not every belief for which we happened to gather, and store, evidence should count as dispositionally justified in the light of such evidence. I may well have acquired, on the 1st of January 1992, evidence for the belief that, on that day, it was raining in Salzburg. However, the belief that on the 1st of January it was raining in Salzburg is hardly D-justified, and if true D-known, unless there are special reasons why that evidence, its temporal distance notwithstanding, is nonetheless readily accessible for me. This last observation seems to show that just a part of one s stored evidence has to be characterized as dispositionally possessed. This raises the important question as to how the boundary has to be traced. I am not going to describe in detail the conditions one piece of stored evidence has to satisfy, for it to count as dispositional evidence. However, the core idea will involve the thought that a subject S can be credited the possession of D-evidence E only if (i) S has already acquired and stored E, and (ii) if E satisfies a condition, we might say, of psychological accessibility; where, roughly, a body of evidence is psychologically

320 T. Piazza accessible iff, given the cognitive creatures we are, it is psychologically possible for us to access it 6. It seems that just the joint satisfaction of (i) and (ii) makes a body of evidence dispositionally available. Once the distinction between E-dispositions and D-evidence is in place, it is apparent that Goldman s counterexample misses its target. The conclusion his argument supports is that having the disposition toward an evidential state (an E- disposition) does not suffice for justification. However, Conee & Feldman s proposal can now be reformulated as involving reference just to the D-evidence one possesses as a consequence of having acquired and stored some evidence. With the use of the notions of D-evidence we can reformulate the evidentialist principle in the following way: (D-j***) If the O-evidence E would justify the belief that q, the D-evidence E D- justifies the belief that q. Goldman s counterexample just establishes that the evidentialist is committed to rejecting the following principle: (G1) If the O-evidence E would justify the belief that p, an E-disposition toward E D-justifies the belief that q As it is apparent, the commitment to reject (G1) does not entail the commitment to reject (D-j***). In a word, Goldman s argument is silent as to whether (D-j***) should be accepted. By the same token, (D-j***) also explains why Alex is not justified to entertain any belief concerning the number of the guests in the restaurant. True, we may suppose that Alex s arithmetical skills, together with the information at his disposal, sustain the counterfactual according to which, were he to consider the question, she would generate the relevant evidence. However, Alex has not yet performed the calculation, and so just possesses a disposition to acquire evidence, not evidence of a dispositional kind. *** If what precedes is on the right track, I have shown, contrary to Goldman, that there is a suitable notion of dispositional justification the evidentialist can use to vindicate the claim that there can be knowledge of propositions one is not occurrently entertaining. Quite naturally, the proposal would seem to involve the thought that a belief which a subject S does not occurrently entertain amounts to D- knowledge just provided that the belief is true and D-justified, namely just provided that S is in possession of D-evidence which the belief that p fits. However, this conclusion would be a little too hasty: as Conee and Feldman perfectly realize, in fact, S knows that p cannot in general be analyzed as S has a true belief which fits the available evidence ; so that, for the same sort of reasons, also S D-knows that p cannot simply be analyzed as S has a true D-belief which fits the available D-evidence. To see why the general characterization does not work, just consider 6 It is arguable that also the satisfaction of (ii) just provides a necessary condition for the dispositional possession of a body of evidence. Given average mathematical skills, Tom s belief that 650 9 362 equals 235,300 is psychologically accessible, though it does not count as evidence Tom possesses.

Evidentialism and the problem of stored beliefs 321 that a subject S may form a true belief which fits the available evidence E and yet fails to be know, just because S forms the belief for the wrong sort of reasons, by relying on aspects of E which do not really bear on the epistemic status of the belief that p. Just imagine the following situation: although Thomas is in possession of good evidence which the belief that his son Arturo will be born around the beginning of April fits evidence delivered by an echograph, medical pronouncements, etc. he starts believing that Arthur will indeed be born in that period because in that period the moon is full, and he believes that a full moon makes a birth likelier. Now suppose that Thomas was actually right, and that Arturo is born the 3rd of April. Does this former belief of his deserve the honorific title of knowledge? I guess that the intuitive and correct answer is a negative one, because this belief, although one that used to fit the evidence available to Thomas, was formed on the basis of a different body of bad evidence. These considerations actually motivate the requirement that a true belief, above and beyond fitting the available evidence, be also formed on the basis of such evidence if it is to count as knowledge. In Conee and Feldman s terminology, the requirement is that the belief be well-founded. When applied to the particular notion of D-knowledge, these considerations seem to suggest the following amended characterization: (D-k) The true proposition that p is D-known if (i) a subject dispositionally believes that p, (ii) the belief that p is dispositionally well-founded. Clearly in order to understand (D-k) it is necessary to understand what does it mean, for a dispositional (stored) belief, to be dispositionally well-founded. As we shall see in this final part, the answer quite naturally flaws from the characterization of the notion of an (occurrently) well-founded belief, with just two additional conditions needed to adjust the occurrent sense to the peculiar features of the dispositional realm. As an O-belief is well-founded if it is based on supporting evidence, a natural proposal would seem to involve the thought that a stored belief is dispositionally well-founded if and only if, at the time of its formation, it was based on supporting evidence. However, one crucial asymmetry between occurrent and dispositional beliefs seems to motivate, in the latter case, two requirements which need not be explicit in the former. An O-belief, on the one hand, is well-founded if it is based on evidence which the belief fits. So there is no questioning the epistemic status of a well-founded belief: any such belief, having been formed on the basis of supporting evidence, is a fortiori a belief which is supported by this evidence; in other words, it does fit this evidence. Therefore, in its occurrent sense, that a belief is well-founded entails that it is justified. On the other hand, this transition does not seem to be guaranteed in the dispositional case. If a stored belief is dispositionally wellfounded, in the advertised sense in accordance to which it has merely been formed on the basis of supporting evidence, it may still fail to be (dispositionally) justified. For it constitutes a possibility that the supporting evidence E, on the basis of which a belief has been formed, as time goes by be lost or forgotten, or have become unserviceable. This possibility clearly evaporates if a belief, above and beyond having been formed on the basis of supporting evidence E, is actually D-justified on

322 T. Piazza the basis of E; if, that is to say, this evidence is dispositionally available to the subject. The satisfaction of this additional condition arguably suffice to prevent the undesired situation of a well-founded and less-than-justified belief from arising: if a belief fits D-evidence E, such evidence can be neither lost, nor forgotten, nor have turned into unserviceable evidence. Accordingly, the following characterization of dispositional well-foundedness seems more apt to qualify that property which may turn a true stored belief into D-knowledge: (W-FD) A stored belief is dispositionally well-founded if (i) it has been formed on the basis of supporting evidence E, on the basis of which (ii) the belief is D-justified. This is not yet enough, though. A second requirement, also stemming from the advertised asymmetry, concerns the dispositional ersatz of the notion of a subject s using a particular body of evidence to form a belief. The problem is that by the mere fact that a belief that p is formed at some t by using supporting evidence E, and, as time goes by and the belief is stored, by the fact that this evidence turns into D-evidence which the belief fits, it does not seem to follow that the subject also possesses the disposition to use E to justify her belief that p whenever it is reactivated in response to circumstances bearing onto whether p. However, for a body of evidence E to be one on the basis of which a stored belief that p is dispositionally well-founded that is to say, if true one which constitutes dispositional knowledge it seems necessary that, above and beyond being supporting evidence kept dispositionally available, it also be evidence which the subject retains the disposition to use under the appropriate circumstances. This leads to the supposition that if S s belief that p, att, is well-founded, in that it is based on supporting evidence E which is thereafter kept dispositionally available, the stored belief that p, for any t n later that t, is dispositionally well-founded only if S would occurrently believe that p on the basis of E were the question whether p to arise. If we dub the relation which obtains between a dispositional belief and a piece of dispositional evidence E when the latter counterfactual condition is satisfied counterfactual well-foundedness (W-F C ), in order to distinguish it from actual wellfoundedness (W-F A ) which is the relation a dispositional belief bears to a piece of supporting evidence E if such belief has actually been formed by relying on E, all this leads to the requirement that a stored belief be also counterfactually wellfounded if it is a belief a subject dispositionally knows. This leads to the following amended characterization: (W-FD) S s stored belief that p is dispositionally well-founded iff (i) it has been formed on the basis of supporting evidence E, on the basis of which (ii) the belief is D-justified, and (iii) the belief that p is counterfactually wellfounded with respect to E. Putting together all these considerations, the final result seems to be the following more perspicuous characterization of D-knowledge: (D-k*) The true proposition that p is D-known iff (i) a subject dispositionally believes that p, and (ii) the belief that p is well-founded D.

Evidentialism and the problem of stored beliefs 323 By resorting to (D-k*) the evidentialist can claim many true stored beliefs to be known, independently of their failing to be supported by evidence a subject occurrently entertains. For it seems quite natural to conjecture that many stored beliefs have been acquired by resorting to supporting evidence which has thereby been kept dispositionally available, with the consequence that the subject of those beliefs has also acquired a disposition to reactivate and use that evidence when her practical needs require a full deployment of distinct items of her knowledge. And any such deployment is hardly to be regarded as the manifestation of something that wasn t already there. *** In this paper I have addressed the problem of the justificatory status of stored beliefs from an evidentialist perspective. Contrary to Goldman, I have denied that the evidentialist is unable to account for the justificatory status it is intuitive to attribute to many (stored) beliefs that are not accompanied or sustained by occurrently entertained evidence. Following Feldman s suggestion I have defended the view that such beliefs are justified dispositionally. As to the notion of dispositional justification, I have shown that Goldman s counterexample draws on an incorrect picture of Conee and Feldman s commitments. However, I have also stressed some problems in alternative characterizations of dispositional justification that can be evinced by Conee and Feldman s writings. The major shortcoming seems to be that the identification of dispositional evidence with stored evidence seems to lack of explanatory value in the absence of a clear criterion that explains why, from an evidentialist point of view, stored evidence should count as available in some sense. I have propose to identify such criterion by generalizing Audi s distinction between D-beliefs and B-dispositions, and by maintaining that stored evidence is epistemically relevant because it is evidence one possesses dispositionally. As to the notion of dispositional knowledge, which the evidentialist needs to characterize the epistemic status of stored beliefs intuitively amounting to knowledge, I have suggested that it should incorporate the extra-condition introduced by Conee and Feldman according to which a true belief must be well-founded, above and beyond fitting the available evidence, if it is to constitute knowledge. As the notion of well-foundedness is normally introduced as a property of occurrent beliefs, however, I have dwelt a little to explain the idea of a dispositionally well-founded belief. According to my proposal, a stored belief isn t dispositionally well-founded if a subject, above and beyond having formed the belief on the basis of supporting evidence, has not the disposition to use the D-evidence such belief fits in order to justify the belief, once occurrently entertained. Acknowledgments I am very grateful to many persons for valuable discussions and important comments on previous versions of this paper. In particular I would like to thank Luca Moretti, Johannes Brandl, Wolfgang Huemer, Gerhard Schurz, and the people from the 2006 Meeting of the Canadian Society for Epistemology. A special thank to Richard Feldman. Finally I am indebted to an anonymous referee for their insightful criticisms. References Audi, R. (1994). Dispositional belief and dispositions to believe. Noûs, 28, 419 434. Audi, R. (1998). Epistemology. New York & London: Routledge.

324 T. Piazza Conee, E., & Feldman, R. (1985). Evidentialism. Philosophical Studies, 48, 15 34 (Reprinted in E. Conee & R. Feldman 2004, with a Forward, 83 107). Conee, E., & Feldman, F. (2001). Internalism defended. In H. Kornblith (Ed.), Epistemology: Internalism and externalism (pp. 231 260). London: Blackwell. Conee, E., & Feldman, R. (2004). Evidentialism. Oxford: Oxford University Press. Feldman, R. (1988). Having evidence. In D. F. Austin (Ed.), Philosophical analysis (pp. 83 102). Dordrecht: Kluwer. Ginet, C. (1975). Knowledge, perception, and memory. Dordrecht: Reidel. Goldman, A. (1988). Strong and weak justification. In A. Goldman (Ed.), 1992, Liaisons. Philosophy Meets the Cognitive Sciences (pp. 127 142). Cambridge (Mass.): MIT Press. Goldman, A. (1999). Internalism expoused. The Journal of Philosophy, 96, 271 293. Goldman, A. (2002). Pathways to knowledge. Oxford: Oxford University Press.