Internalism Defended Author(s): Richard Feldman and Earl Conee Source: American Philosophical Quarterly, Vol. 38, No. 1 (Jan., 2001), pp. 1-18 Published by: University of Illinois Press on behalf of North American Philosophical Publications Stable URL: http://www.jstor.org/stable/20010019 Accessed: 08/01/2009 17:48

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=illinois.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

University of Illinois Press is collaborating with JSTOR to digitize, preserve and extend access to American Philosophical Quarterly.

http://www.jstor.org American Philosophical Quarterly Volume 38, Number 1, January 2001

INTERNALISM DEFENDED

Richard Feldman and Earl Conee

Anternalism in has been get? the terms are routinely applied to theories ting bad press lately. Externalism is that predate their use. Thus, many propo? ascendant, partly because insurmountable nents of theories of justification have not problems for internalism are supposed to classified their views as internalist or have been identified.1 We oppose this trend. externalist. The recent literature is, there? In our view the purported problems pose fore, the best source of information about no serious threat, and a convincing argu? the nature of the distinction. Here are a few ment for internalism is untouched by the examples of how internalism has been recent criticism. identified. Laurence BonJour writes: Our main here is to refute goal objections most The generally accepted account... is to internalism. We what begin by offering that a theory of justification is internalist if we think is the best to understand the way and only if it requires that all of the factors distinction between exter? internalism and needed for a to be epistemically justi? nalism. We then present a new argument fied for a given person be cognitively for internalism. This frees internalism from accessible to that person, internal to his cog? nitive what we regard as suspect deontological perspective.2 underpinnings. Finally, we reply to what writes: we take to be the most significant objec? Some examples suggest that justification is tions to internalism. grounded entirely in what is internal to the mind, in a sense that it is acces? I.What is Internalism? implying sible to introspection or reflection by the we Internalism and externalism are views subject?a view might call internalism about about which states, events, and conditions justification.3 can contribute to epistemic justification? writes: the sort of that, in sufficient justification The basic thrust of internalism in epistemol is a condition for strength, necessary ogy, therefore, is that the properties that Use of the terms "internalist" . confer warrant upon a belief are properties and "externalist" to classify theories of to which the believer has some special sort justification is a recent development, and of epistemic access.4

1 2 / AMERICAN PHILOSOPHICALQUARTERLY

John Pollock writes that: the special kind of access on which many internalist theories can reach Internalism in epistemology is the view that rely only mental items, and all mental items, only internal states of the cognizer can be rel? perhaps or at least all that be counted as evant in determining which of the cognizer's might play? a beliefs are justified.5 ing role in justification. We think that and clarity are Finally, characterizes one ver? best served by understanding internalism sion of internalism this way: as mentalism. "Internalism" is a recent Justification requires only really proper technical term. It has been introduced to thought on the part of the subject: if a be? refer to a variety of theories in epistemol liever has obtained and sustains his belief ogy that share some vaguely defined salient through wholly appropriate thought, then the feature. Any definition of the term is to believer is justified in so believing?where some extent Mentalism codi? the appropriateness of the thought is a mat? stipulative. fies one standard in which the word ter purely internal to the mind of the subject, way has been used. and not dependent on what lies beyond.6 Somewhat more internalism as We find two distinct but related precisely, closely we characterize it is committed to the fol? characterizations of internalism in passages two theses. The first asserts the such as these. One characterization uses a lowing strong of notion of access. What we shall call supervenience epistemic justifi? cation on the mental: "accessibilism" holds that the epistemic S. The status of a justification of a person's belief is deter? justificatory person's doxastic attitudes strongly supervenes on mined by things to which the person has the person's occurrent and dispositional some special sort of access. BonJour calls mental states, events, and conditions. this access a "suitable awareness."7 Audi thesis out a says that the access is through "introspec? The second spells principal tion or reflection." Others say that the implication of S: access must be "direct."8 The quotations M. If any two possible individuals are exactly a from Pollock and Sosa suggest somewhat alike mentally, then they are exactly alike same are different account. They suggest that justificationally, e.g., the beliefs internalism is the view that a person's be? justified for them to the same extent.10 liefs are things that are justified only by (M) implies that mental duplicates in differ? internal to the mental life. We person's ent possible worlds have the same attitudes shall call this version of internalism "men justified for them. This cross world compari? talism."9 A mentalist may assert that theory son follows from the strong supervenience is determined oc justification entirely by condition in (S).11 Externalists characteristi? current mental factors, or by dispositional cally hold that differences in justification can ones as well. As as the that are long things result from contingent non-mental differ? said to contribute to are in the justification ences, such as differing causal connections mind, the view as a ver? person's qualifies or reliability. Theories that appeal to such sion of mentalism. factors clearly deny (S) and (M). Thus, our We think it that have likely philosophers way of spelling out the internalism/exter not mentalism from accessibilism separated nalism distinction properly classifies because have assumed that the they tacitly characteristically externalist views. extensions of the two do not differ in any One advantage of our way of understand? way. have assumed that significant They ing the distinction between internalism and INTERNALISMDEFENDED / 3

one externalism in epistemology is that it Thus, modest asset of mentalism is closely parallels the counterpart distinction that it renders readily intelligible the nomi? to in the of mind.12 In the philoso? nal connection of epistemic internalism phy of mind case, the main idea is to mind internalism. A much stronger consid? that it turns distinguish the view that the contents of eration in favor of mentalism is as we attitudes depend entirely on things within out to be entirely defensible, shall a person's own cognitive apparatus from try to show. the view that there are factors external to IL A Defense of Internalism the person that help to determine attitudinal content. Mind internalism is naturally ren? Our argument for internalism focuses on dered as a supervenience thesis. Roughly, the pairs of examples that we take to be repre? thesis is that a person's mental content su? sentative. Either in one member of the pair pervenes on the person's "purely internal" someone has a justified belief in a propo? states, events, and conditions. The relevant sition while someone else's belief in that as supervenience base cannot be specified is not justified, or one person's "the mental," as we have done for epistemic belief is better justified than the other's. internalism, since a mental states, person's We contend that these contrasts are best events, and conditions are trivially suffi? explained by supposing that internal dif? cient for the person's attitudes with their ferences make the epistemic difference. specific contents. But the root idea is the Here are the examples. same. The mind internalist is trying to ex? Example 1) Bob and Ray are sitting in clude such plainly external factors as the an air-conditioned hotel lobby reading environmental causal origins and the so? yesterday's newspaper. Each has read that cial milieu of the person's attitudes. it will be very warm today and, on that Likewise, the epistemic internalist is prin? basis, each that it is very warm cipally opposed to the existence of any today. Then Bob goes outside and feels the justification-determining role for heat. They both continue to believe that it plainly external factors such as the general is very warm today. But at this point Bob's accuracy of the mechanism that produces belief is better justified. a given belief or the belief's environmen? Comment: Bob's justification for the tal origin. Mentalism bears this out. belief was enhanced by his experience of What internalism in epistemology and phi? feeling the heat, and thus undergoing a losophy of mind have in common is that mental change which, so to speak, "inter? being in some condition that is of philosophi? nalized" the actual temperature. Ray had cal interest?being epistemically justified in only the forecast to rely on. certain attitudes, or having attitudes with Example 2) A novice bird watcher and certain contents?is settled by what goes on an expert are together looking for birds. inside cognitive beings. The condition of in? They both get a good look at a bird in a terest is in this sense an "internal" matter, nearby tree. (In order to avoid irrelevant thus justifying the use of this term. Mental? complexities, assume that their visual pre? ism obviously captures this feature of sentations are exactly alike.) Upon seeing internalism. Accessibilism captures it only the bird, the expert immediately knows that when conjoined with the further thesis that it is a woodpecker. The expert has fully what is relevantly accessible is always inter? reasonable beliefs about what woodpeck? nal to something, presumably the mind.13 ers look like. The novice has no good 4 / AMERICAN PHILOSOPHICALQUARTERLY

reason to a believe that it is woodpecker Smith's reasons derive from Jones pretend? and is not justified in believing that it is. ing to own a Ford. Someone else in the Comment: The epistemic difference be? office, unknown to Smith, does own a Ford. tween novice and expert arises from The fact that Jones is merely simulating something that differentiates the two inter? Ford ownership keeps Smith from know? nally. The expert knows the look of a ing that someone in his office is a Ford woodpecker. The novice would gain the owner, but it does not prevent Smith from same justification as the expert if the novice being justified or diminish his justification. came to a share the expert's internal condition At later time Smith gains ample reason concerning the look of woodpeckers. to believe that Jones is pretending. At that Example 3) A Teaching Assistant point Smith is not justified in believing and a beginning logic student are looking either that Jones owns a Ford or that some? over a homework assignment. One ques? one in his office owns a Ford. tion a sentence displays that they both Comment: Again the epistemic change know to express a and asks whether occurs when a suitable external fact?this certain other sentences are true as well. The time, the fact that what Smith has seen is can TA easily tell through simple reflec? Jones pretending to own a Ford?is tion that some of the other sentences brought into Smith's mind. The difference express logical consequences of the origi? between Smith being justified in believing nal sentence and thus she is justified in that Jones owns a Ford (and that someone believing that they are true as well. The in the office owns a Ford) in the one case student is clueless. and not in the other is an internal change Comment: Again there is an internal dif? in Smith. ference between the two. The difference is Example 5) Hilary is a brain in a vat who that the TA has justification for her beliefs has been abducted recently from a fully to the effect that certain val embodied life in an ordinary environment. idly follow from the original one. She is He is being stimulated so that it seems to expert enough to "see" that the conclusions him as though his normal life has contin? follow without performing any computa? ued. Hilary believes that he ate oatmeal for tions. This case differs from example 2 in breakfast yesterday. His memorial basis for that here the mental difference concerns his breakfast belief is artificial. It has been cognizance of necessary of logic induced by his "envatters." Here are two whereas in example 2 the expert was cog? versions of relevant details. nizant of contingent facts about visual 5a) Hilary's recollection is very faint and characteristics of woodpeckers. But just as lacking in detail. The meal seems incon? in example 2, relevant internal differences gruous to him in that it strikes him as a make the difference. The beginning student distasteful breakfast and he has no idea could come to share the epistemic state of why he would have eaten it. the TA by coming to share the TA's familiar? 5b) Hilary's recollection seems to him ity with the logical consequence relation. to be an ordinary vivid memory of a typi? Example 4) Initially, Smith has excellent cal breakfast for him. reasons to believe that Jones, who works Comment: Although in both (5a) and (5b) in his office, owns a Ford. Smith deduces Hilary's breakfast belief is false and its that someone in the office owns a Ford. The basis is abnormal, the belief is not well latter belief is true, but the former is false. justified in (5a) and it is well justified in INTERNALISMDEFENDED / 5

(5b). Hilary in (5a) differs internally from internal states?typically featuring an ex? Hilary in (5b). His mental states in (5b) perience or another justified belief?that include better evidence for the belief than is suitably related to B. Objections of the he has in (5 a). first sort focus on internal states that are In the first four of these examples the supposed to justify beliefs, arguing that location of a relevant item of informa? there are some justified beliefs for which tion?in the mind of a subject or outside there are no internal justifying states. Ob? of it?makes the epistemic difference. In jections in the second group focus on the the fifth example, a purely internal differ? connections between candidate internal ence is decisive. It is reasonable to justifiers and the beliefs they are supposed generalize from these examples to the con? to justify, arguing that internalists inevita? clusion that every variety of change that bly run into insurmountable difficulties brings about or enhances justification ei? when they attempt to say anything definite ther internalizes an external fact or makes about the nature or status of the connections.

a purely internal difference. It appears that While some internalist theories may have there is no need to appeal to anything trouble dealing with some of these objec? extramental to explain any justificatory tions, there are several internalist approaches difference. These considerations argue for the that can deal adequately with all of them. general internalist thesis that these epistemic We concentrate primarily on two ap? differences have an entirely mental origin. proaches, one that limits justifying states In each case, it is natural to regard the to currently conscious mental states and mental difference as a difference in the one that also includes as potential justifiers evidence that the person has. Variations in whatever is retained in memory. Since the presence or strength of this evidence theories of each sort surmount all of the correspond to the differences in justifica? objections, the internalist approach is in no tion. Our favorite version of internalism, danger of a general refutation.16 evidentialism, asserts that epistemic justi? A. Are There Enough Internal Justifiers? fication is entirely a matter of evidence.14 However, our goal here is to defend Al. Impulsional Evidence internalism generally, and not just its Alvin Plantinga's objection focuses on evidentialist version. evidentialist versions of internalism.17 We have no proof that there is no excep? But the same sort of objection seems tion to the pattern exhibited by our equally applicable against any prima facie examples. The argument does not establish plausible internalist view. Plantinga asserts that internalism is true. It does support that there are three views evidentialists internalism. Further support will emerge can hold concerning what constitutes from successful replies to objections.15 evidence, and he argues that each view renders evidentialism III. Objections and Replies unsatisfactory. The three possibilities are: (1) evidence The objections we shall consider fall into consists only of other beliefs (all evi? two broad and overlapping categories. One dence is propositional); (2) evidence sensible general description of internalist consists only of beliefs and sensory theories is that they say belief B is justified states (all evidence is propositional or can just in case there is some combination of sensory); (3) evidence also include 6 / AMERICAN PHILOSOPHICALQUARTERLY

the sense of conviction or confidence impulsional, however, he would be mis? that accompanies beliefs (all evidence is taken in thinking that all beliefs have any propositional, sensory, or impulsional). similar sort of evidential support. There are Plantinga uses knowledge of simple ar? several internal states to distinguish here. ithmetical facts to defend his objection. He Perhaps we feel attracted to the proposi? asserts that we do not believe that 2+1=3 tion that 2+1=3 and we feel impelled to on the basis of propositional or sensory believe it. Not everything we believe feels evidence. So, if evidentialists adopt alter? attractive in this way or any other. For in? natives (1) or (2), their theory implies that stance, some known propositions are this belief is not justified. Yet, of course, believed reluctantly, on the basis of rea? we do know that 2+1=3. Plantinga claims sons, in spite of their seeming distinctly there is a "felt attractiveness" about the unattractive and implausible. Some beliefs content of that belief, and he says 2+1=5 result from fears. They need not seem in "feels wrong, weird, absurd, eminently re any way attractive. Correspondingly, the jectable."18 He calls the "felt attractiveness" denials of things we believe do not always an "impulse" and classifies it as "impulsional feel "weird" or "absurd," even if we think evidence." So internalists might take that they are false. There may be a sense Plantinga's third alternative and claim that of obviousness that accompanies belief in this impulsional evidence is the internal fac? some propositions. This sense may contrib? tor that justifies simple mathematical beliefs. ute to their evidential support. But quite Plantinga argues that there is a problem plainly not all believed propositions share with this account. He claims that necessarily that feature, or anything that resembles it. all beliefs would have similar justification: So it is not true that there is "impulsional "You have impulsional evidence for p just in evidence" for every believed proposition. ... virtue of believing p. It isn't even pos? Furthermore, even if there were sible that you believe p but lack impulsional impulsional evidence for each belief, it evidence for it: how could it be that you would not follow that each belief satisfies believe p although it does not seem to you any plausible evidential version of the jus? to be true?"19 He infers that on this view of tification condition for knowledge. The evidence, the internalist justification con? existence of a bit of supporting evidence dition for knowledge that consists in is clearly not enough. A plausible eviden? having evidence is implied by the belief tial condition for knowledge requires condition. If Plantinga is right about this, something more, such as strong evidence then evidentialists who take alternative (3) on balance, or at least evidence undefeated are stuck with the unacceptable conclusion by other evidence. An impulse to believe that all actual beliefs are justified. The would not always qualify as strong evi? other initially plausible internalist views, dence on balance, or undefeated evidence. for instance, those that appeal to epistemic Moreover, even if there were some responsibility as the key to a belief's justi? impulsional evidence for all beliefs, it fication, seem equally susceptible to this would not follow that all beliefs are justi? sort of objection. The "felt attractiveness" fied to any degree. In some cases anything seems equally to render believing the like impulsional evidence is decisively epistemically responsible course of action to outweighed by competing evidence. There? take. So, again, all beliefs would be justified. fore, the existence of impulsional evidence Even if Plantinga were right in claiming for all beliefs would not render redun? that the evidence for beliefs like 2+1=3 is dant a plausible evidential condition on INTERNALISMDEFENDED / 7

common knowledge and would not saddle internalists personal facts, facts that constitute areas with the unacceptable result that all beliefs knowledge, facts in our of expertise, are justified. and so on. Since we know all these things, Even with regard to the simplest of math? we believe them. These are stored beliefs, ematical beliefs, impulsional evidence of not occurrent beliefs. Since we know them, the sort Plantinga mentions is not our only we are justified in believing them. But on evidence. We have evidence about our suc? what internalist basis can these beliefs be cess in dealing with simple arithmetical justified? As Goldman says, "No percep? matters and knowledge of the acceptance tual experience, no conscious memory our assertions about these matters enjoy. , and no premises consciously enter? So, we have reason to think that our sponta? tained at the selected moment will be neous judgments about simple mathematical justificationally sufficient for such a be? matters are correct. Furthermore, we know lief."21 Internalists are stuck with the that we learned these sorts of things as chil? unacceptable result that these beliefs are dren and we have not had our more recent not justified, unless something internal that assertions about them contradicted by oth? justifies them can be found. ers. If we had been making mistakes about In formulating this objection Goldman these kinds of things, it is very likely that assumes two propositions, either of which problems would have arisen and we would internalists can sensibly reject. On the one have been corrected. Finally, at least ac? hand, he assumes that virtually all justi? cording to some plausible views, we have fied beliefs are stored beliefs. On the other a kind of a priori insight that enables us to hand, he assumes that internalists must find grasp simple mathematical propositions. something conscious to serve as their jus? This insight provides us with some evi? tification. But internalists have good reason dence for the truth of simple mathematical to reject this pair of propositions. One al? truths. Much of this evidence is retained ternative is to argue that, in the most central in memory; some of it is conscious when? sense, few beliefs are justified, and typi? ever such propositions are consciously cally the ones that are justified are apprehended. There seems to be plenty of occurrent. The second option is to argue additional evidence, whether or not that other non-occurrent internal states can justifiers are restricted to conscious states. contribute to the justification of non-oc? Indeed, the suggestion that the only eviden? current beliefs. tial bases for simple arithmetical beliefs are The first response relies on the idea that impulses to believe is extremely implausible. there are occurrent and dispositional senses Thus, Plantinga's objection makes no of "justified," just as there are occurrent real trouble for evidentialism. Any other and dispositional senses of "belief." In the reasonable internalist view clearly has a most fundamental sense of "justified," a similar response available to the counter? belief can be justified for a person only by part objections. the person's current evidence, and one's current evidence is all conscious. In this A2. Stored Beliefs sense, non-occurrent beliefs are typically argues that internal states not justified. However, in the same way cannot account for the of justification that there are propositions in which one has stored beliefs.20 The problem is this. At any stored belief, one can have "stored justifi? moment almost of what we given nothing cations" for these beliefs. That is, one can know is consciously considered. We know have in memory reasons that justify the 8 / AMERICAN PHILOSOPHICALQUARTERLY

are belief.22 Beliefs like this dispositionally or other memories. These stored justifications are justified.23 Thus, although stored beliefs are internalist by the standard of M and they seldom justified in the most fundamental are plausibly regarded as evidence that the sense, they are often dispositionally justi? person has.26 fied. The description presented here of the a Goldman objects to proposal along these second internalist approach leaves open lines that one of us made previously.24 He important questions about which stored takes the general idea behind the proposal to internal states can justify beliefs and what be that a disposition to generate a conscious relation these stored states must have to a evidential state counts as a justifier. He then belief to justify it.No doubt these are dif? raises the following objection: ficult questions. Versions of internalism will differ which stored states Suppose a train passenger awakes from a nap concerning count as But there is no but has not yet opened his eyes. Is he justi? they justifiers.27 that re? fied in believing propositions about the appearance internalism lacks the details of the neighboring landscape? Surely sources to provide satisfactory answers to not. Yet he is disposed, merely by opening these questions. to conscious mental states his eyes, generate However, one might think that external that would such beliefs.25 occurrently justify factors having to do with the actual source of a memory belief can affect its justifica? The idea behind the current proposal is not tion. In fact, Goldman himself describes what Goldman criticizes here. It is not that something similar to our second internalist any conscious mental state that one is dis? approach and claims that it fails for just this posed to be in counts as evidence. The idea reason.28 We turn next to this is that some non-occurrent states that one objection. is such as non-occurrent memo? already in, A3. Forgotten Evidence ries of are stored perceptual experiences, Several authors have raised objections evidence. this stored evi? Presently having involving forgotten evidence.29 We will dence some non justifies dispositionally focus on an example Goldman provides: occurrent beliefs that one already has. The Last year Sally read about the health ben? train does not have the evidence passenger efits of broccoli in a New York Times that he would have received were he to science-section story. She then justifiably state open his eyes. The dispositional that formed a belief in broccoli's beneficial ef? he is in, his disposition to see the landscape fects. She still retains this belief but no his is not stored evidence by opening eyes, longer recalls her original evidential source for propositions about the landscape. It is (and has never encountered either corrobo? or a potential to acquire evidence, and that is rating undermining sources). Nonetheless, her broccoli belief is still if crucially different. justified, and, as a case true, qualifies of knowledge.30 The second solution to the problem of stored beliefs does not invoke a distinction This example illustrates something that must be conceded to be common. We now between occurrent and dispositional justifi? know for which we have cation. Internalists can plausibly claim that things forgotten our evidence. The for if we have numerous ordinary justified be? original problem internalism arises most if we as? liefs that we are not consciously considering, clearly then there is no reason to exclude from what sume that Sally's original evidence is lost and not of stored justifies these beliefs further stored beliefs irretrievably part any INTERNALISMDEFENDED / 9 justification that Sally might have. Let us is not her original evidence. If Sally lacks assume that Sally is occurrently entertain? any supporting background information ing her justified belief about broccoli and and also lacks any reason to trust her that the facts about the original source of memory, then we doubt that her belief the belief are not part of any internalist about broccoli really is justified. justification of it. Externalists might argue Goldman considers and rejects this sec? that the contingent merits of the external ond response on the basis of a new version source of this belief account for its justifi? of the example about Sally.31 The crucial cation. How can internalists explain why feature of the revised example is that the this belief is currently justified? belief originally came from a disreputable One internalist answer to this question source. Sally has the same belief about is that Sally's justification consists in con? broccoli and the same background beliefs scious qualities of the recollection, such as about the reliability of her relevant capaci? its vivacity and her associated feeling of ties. But now it is part of the story that Sally confidence. We see no fatal flaw in this obtained the belief about broccoli from an response. It will be most attractive to article in the National Inquirer, a source internalists who hold that only what is con? Goldman assumes to be unreliable. Goldman scious can justify a belief. We note that not claims that all memory beliefs are justified according Sally cannot be credited with justifiably be? to this Some beliefs are theory. memory lieving that broccoli is healthful. Her past a sense of and accompanied by uncertainty acquisition is still relevant, and decisive. At a lack of confidence. Other memory beliefs least it is relevant so long as we are consider? are accompanied by a recognition of com? ing the "epistemizing" sense of justification, in which carries a true belief a peting evidence. This competing evidence justification distance toward be? can render vivacious memory beliefs unjus? good knowledge. Sally's lief in the healthfulness of broccoli is not tified. These are plausible results, so this in that for she does not restrictive version of internalism does have justified sense, surely know that broccoli is healthful given that the the resources to deal with forgotten evidence. National was her sole source of in? Another defensible answer is available Inquirer formation.32 to internalists who think that not all evi? We that does not know that dence is conscious. If Sally is a normal agree Sally broccoli is healthful under these condi? contemporary adult, she is likely to have tions. We also that facts about her quite of a bit of readily retrievable evidence agree of the belief determine this re? supporting her belief about broccoli. The acquisition sult. However, it does not follow that healthfulness of vegetables is widely re? Sally's belief is not The ported and widely discussed. Furthermore, justified. "epistemizing" sense of is her belief about broccoli is probably not justification said by Goldman to be a sense to which a belief undermined by any background beliefs she according that is is one that has been carried is likely to have. Finally, she, like most justified "a distance toward This people, probably has supporting evidence good knowledge." fits with our initial characterization of consisting in stored beliefs about the gen? as the sort which is eral reliability and accuracy of memory. epistemic justification for But from the fact She knows that she is generally right about necessary knowledge. that belief falls short of this sort ofthing. So Sally would have jus? Sally's knowledge, it not not tification for her broccoli belief, though it does follow that it has been carried 10 / AMERICAN PHILOSOPHICALQUARTERLY

a good distance toward knowledge. Thus, present Sally case, the otherwise generally an initial weakness in this objection is that good credentials of her memorial beliefs sup? its concluding inference is invalid. port the belief and are undefeated. A second fault is that the allegedly un? Some confirmation of our analysis comes justified belief is actually a justified true from comparing the case as described to a belief that is not knowledge. It is a Gettier case in which Sally does remember the case. We endorse the following rule of unreliability of her source but retains the thumb for classifying examples of true be? belief anyway. It is clear that there would liefs that are not knowledge: be something far less reasonable about her belief in that situation. This that RT. If a true belief is accidentally correct, suggests the source does make the belief in spite of its being quite reasonably be? forgetting lieved, then the example is a Gettier case. better justified. Further confirmation emerges from con? RT helps to show that the second version trasting the example with yet another of the example about Sally is a Gettier case. variation. Suppose Sally believes both that Sally believes that broccoli is healthful. broccoli is healthful and that peas are She believes (presumably justifiably) that healthful. that her source for the she learned this from a reliable source. She Suppose former is still the National Inquirer but her is wrong about her source but, coinciden source for the latter belief is the reliable tally, right about broccoli. This fits exactly New York Times. Again she has forgotten the pattern of Gettier cases, and RT classi? her sources, but she correctly and reason? fies it as such. It is a quite reasonable belief ably believes that she virtually always gets on Sally's part which, in light of its unreli? beliefs like these from trustworthy sources. able source, is just accidentally correct. It is Goldman's objection requires differentiat? a justified true belief that is not knowledge. ing these two beliefs in an unacceptable Our view has an implication that may way. It counts the former belief as unjusti? initially seem odd. When Sally first came fied, on the basis of the unreliability of its to believe that broccoli is healthful, the forgotten source. Yet from Sally's present belief was unjustified because Sally had the two are on a reason to distrust her source. Yet we seem perspective, propositions par. It would be completely unreasonable for her in effect to be saying that merely because to give up one belief but not the other. The she has forgotten about that bad source, the best thing to say is that both are justified, but belief has become justified. We are not the broccoli belief does not count as knowl? quite saying that. As we see it, when she edge because it is a Gettier case. forgets about the source she has lost a We conclude that internalism does not defeater of a justification for her broccoli have any difficulty finding adequate justi? belief. Assuming that Sally knows herself fication in cases of forgotten evidence.33 normally to be judicious about her sources, any belief she retains thereby has consid? B. Links and Connections erable internal Whatever beliefs support. We turn next to two objections concern? she retains are this, unless justified by they ing the connections between perceptual are defeated. A belief is defeated in case any experiences or other justified beliefs and in which she has indications that impeach the beliefs they are supposed to justify. what it is reasonable for her to take to be the There are difficult questions about exactly source of her belief. But when she no longer how these states manage to justify the be? such indication, as in the possesses any liefs they support. These are problems of INTERNALISMDEFENDED / 11

detail, and internalists have reasonable matter of conforming to duties one must be are choices concerning how to work out the in a position to know about. Internalists details. As we shall show by responding to free to reject that conception.35 They need not several related objections, there are no defend an identification of justification with not unresolvable problems here. any sort of duty fulfillment. They need defend anything that makes having justified Bl. The Need Order for Higher Beliefs beliefs depend on having some way to know has that the con? argued what justifies what. To cite our favorite in? siderations that internalism support equally stance, evidentialists hold that the possession the of what he calls a support imposition of the right evidence by itself secures the order on "higher requirement" justifica? justification of the corresponding beliefs. tion. The idea is that if the argument that The justification supervenes on the inter? leads to the conclusion that internal only nal possession of appropriate evidence. factors can serve as is sound, then justifiers Neither epistemic evaluations nor duties a to the con? there is also sound argument need enter in at all. clusion that for a belief to be the justified It might be thought that evidentialism believer must be able to tell which factors should be formulated in ways that require the belief. Alston writes: justify for justification not only supporting evi? Suppose that the sorts of things that can count dence but also knowledge of higher level as are to justifiers always accessible me, but principles about the justificatory efficacy that it is not accessible to me which items of of this evidence. Some internalists do seem these sorts count as for which justifications to impose such a requirement.36 We agree beliefs. I have access to the justifiers but not with Alston that any such theory is implau? to their efficacy. This will take justificatory sible, implying that few people have away my ability to do what I am said to have justified beliefs. However, we see no reason an obligation to do just as surely as the lack to think that evidentialists, or internalists of access to the justifiers themselves. To il? must can generally, endorse any higher order lustrate, let's suppose that experiences evidence can make function as justifiers, and that they are ac? requirement. Having for on its own. cessible to us. I can always tell what sensory justification experiences I am having at a given moment. The appearance that justifying relations Even so, if I am unable to tell what belief pose a problem for internalism arises partly current environment is about the physical from formulating the debate between justified by a given sensory experience, I am internalists and externalists as a debate unable to my thereby regulate perceptual over whether all "justifiers" are internal. beliefs according as they possess or lack ex? For example, Goldman takes to internalists periential justification.34 to require that all "justifiers" must be in Alston goes on to argue that this higher level some suitable way accessible.37 This way one us are requirement is that few of able to of formulating the issue is problematic. satisfy, and he rejects the requirement partly Suppose that a person who believes q on for this reason. Since the argument for the the basis of believing p is justified in be? higher order requirement is clearly unsound, lieving q. We might then say, as a first Alston concludes that the original argument approximation, that the justifiers for q are for internalism is unsound as well. (i) the belief that p together with its justi? The argument that Alston is considering fication, and (ii) the fact that p justifies q. relies on a deontological conception of justi? The fact in (ii) is not itself an internal state, to a fication according which justification is and so itmight be thought that internalists 12 / AMERICAN PHILOSOPHICALQUARTERLY

are faced with the difficult task of finding further that q is an extremely simple and in? some internal representation of this state tuitively obvious (to us) logical consequence to serve as a justifier.38 of p. For q to be justified for the person, must a sense There is in which p's support for he have additional evidence, analogous the a an q is "justifier." It is part of explana? TA's additional evidence, for the proposition tion of the fact that the person's belief in q that q follows from p? is justified. But this does not imply that One possible view is that the answer is internalists are committed to the view that "No". According to this view, there are some there must be internal representation certain elementary logical connections that a of this fact. Itmay be that person's being are necessarily reflected in epistemic con? in the state described by (i) is sufficient nections. The best candidates for this for the belief that q to be justified. If so, relation include cases where one proposi? then all individuals mentally alike in that tion is a conjunction of which the other is they share that state are justified in believ? a conjunct. The general idea is that some ing q. The fact in (ii) may help to account propositions, p and q, have a primitive or for the justification without the person basic epistemic connection. If p and q have making any mental use of that fact. this connection, then, necessarily, if a per? General beliefs that relate evidence to a son is justified in believing p, then the a conclusion sometimes do make justifica? person is also justified in believing q. Per? some tory difference. This occurs in of the haps it is part of understanding p that one examples in our argument for internalism. grasps the connection between p and q. But the sort of connecting information that There is, then, no need for additional in? the examples suggest to be necessary is non formation about the link between p and q epistemic information that justified believers that a person who is justified in believing typically have. The logic TA, for example, p might lack. By the test of the superve? had justification for beliefs about implica? nience thesis asserted by S, internalists can tion relations that the student lacked. The accept this answer. expert bird watcher had justification for be? Internalists can also hold that the answer liefs about what woodpeckers look like. This to the question above is "Yes." In this case, might take the form of various generaliza? there is something resembling a higher or? tions, e.g., any bird that looks like that is a der requirement. However, it is not any woodpecker, any bird with that sort of bill is implausible requirement to the effect that one a woodpecker, etc.39 The student and the nov? have beliefs about justification. It is merely ice bird watcher lacked these justifications. a requirement that one have evidence that Itwould be amistake, however, to argue from there is a supporting connection?for instance, these cases to any universal "higher order the logical consequence relation?between requirement," especially to a higher order what is ordinarily regarded as one's evidence requirement to have epistemic information and what it is evidence for. This evidence can about what justifies what. come from direct insight or from any other A fully developed internalist theory must source. This is evidence that people normally state whether linking information of the sort have in a variety normal situations.40 possessed by the logic TA and the expert bird A similar question arises concerning watcher is required in the case of simpler perceptual beliefs about the qualities of connections. Suppose that a person has a jus? the objects one is perceiving. We said tified belief in some proposition, p. Suppose above that the expert bird watcher has INTERNALISMDEFENDED / 13 background information about the look introspective belief that one is having that of woodpeckers that justified the belief experience, while our experience of a 23 that he saw a woodpecker. The novice sided figure does not justify for us the belief lacked that information. The new ques? that we are having that experience? tion concerns simpler qualities such as Internalism has resources to explain why redness. Must a person with a clear view the two experiences have different epistemic of a red object have evidence about the consequences. We can best explain the rel? look of red things in order to be justi? evant internal features through consideration fied in believing that there is something of some hypothetical person who does have red before him or is the mere experience the ability to identify 23-sided figures in his of redness (in the absence of defeaters) visual field and contrasting this person with sufficient for justification? ordinary people who lack that ability. Accord? Again, it is not crucial to answer this ques? ing to one internalist option, someone who tion here. What is important for present has the ability has an experience qualitatively purposes is that internalists have plausible different from the experiences of those who options. If an experience of the phenomenal lack that ability. We will call the quality that quality corresponding to redness automati? underlies the ability "recognition." It can cally justifies thebelief (absent any defeater), plausibly be held that recognition makes a then people internally alike in that they share justificatory difference. When our visual field the experience will be justified in believing contains a triangle that contrasts clearly with the same external world proposition. If in? its surroundings, we recognize it as such. We formation about the look of red objects is do not similarly recognize 23-sided figures. required, then people internally alike in that The recognition is not a true belief linking they share this information as well as the ex? the experience to a belief about its content.42 perience of red will have the same external It is, instead, a feature of experience itself. world proposition justified. There is a prob? This experiential feature is what makes it true lem for internalism here only if there is some that triangles, optimally viewed, are gener? reason are to think that internal differences ally seen as triangles, while 23-sided figures, inadequate to account for some difference in even when optimally viewed, are not gener? justification. We see no threat of that. ally seen as being 23-sided. It is this aspect of the experience that provides evidential B2. Justification of Introspective Beliefs support for the corresponding belief. For Ernest Sosa raises a problem about how most of us, this sort of feature is present when experiences justify introspective beliefs: we experience clearly discriminable triangles a Some experiences in certain sense 'directly and not present when we experience 23-sided fit' some beliefs. But not all ex? introspective figures. But a person who did have that re? fit the beliefs periences directly introspective markable ability would have an experience that describe them Thus my belief correctly. qualitatively unlike ours. that at the center of my visual field there lies a Rather than appealing to any qualitative white triangle against a black background difference in experience, internalists can would so fit the corresponding experience. But appeal instead to background information. my belief that my visual field contains a 23 Ordinary people have learned that the prop? sided white figure against a black background of a three-sided is would not fit that experience.41 erty being image associated with a certain sort of visual ap? The question, then, is this: Why does having pearance. They have not learned which a suitable experience of a triangle justify the sorts of visual appearances are associated 14 / AMERICAN PHILOSOPHICALQUARTERLY

with being a 23-sided image. On this view, IV. Conclusion only by learning some such association could We have defended internalism not a person have justification from experience just to but to move the debate for making these sorts of classifications of praise it, beyond it.We have tried to show that no images. Internalists can plausibly appeal to genuine for this of theories has this sort of background information as the problem category internal difference that accounts for differ? been identified. We have seen that even versions of internalism that on ences in justification in these cases. As in the depend only conscious elements have not been refuted. cases considered in section Bl, the informa? Various less restrictive views about what tion here is not epistemic information about determines have en? what justifies what, information people typi? justification emerged unscathed as well. On account cally lack. It is simply information about tirely any of what internalism the one properties that are associated with experi? is, including we have offered internalism is noth? ences of certain types. here, more than a broad doctrine about the We conclude that Sosa is right to say that ing location of the factors for some but not all experiences lead to justifi? determining that cation of introspective beliefs that correctly epistemic justification. Having argued internalist views stand in no of describe them. But internal differences, ei? jeopardy refuted, we recommend ther in the experiences themselves or in being generally that attention focus on background information, are available to ac? epistemological count for the difference between those that more specific accounts that are more in? formative.43 do lead to justification and those that do not.

University of Rochester

NOTES

This paper is a somewhat abbreviated version of a paper that will appear inEpistemology: Internalism and Externalism, edited by , (Maiden, Mass.: Blackwell, forthcoming).

1. For a summary of the current state of epistemology that illustrates this sort of view, see Philip Kitcher, "The Naturalists Return," The Philosophical Review, vol. 101 (1992), pp. 53-114.

2. Laurence BonJour "Externalism/Internalism," in A Companion to Epistemology, ed. Jonathan Dancy and Ernest Sosa (Oxford: Blackwell, 1992), p. 132.

3. Robert Audi, Epistemology: A Contemporary Introduction to the Theory of Knowledge (New York: Routledge, 1998), pp. 233-234. Emphasis in the original.

4. Alvin Plantinga, Warrant: The Current Debate (Oxford: Oxford University Press, 1993), p. 6. A very similar formulation appears inWilliam Harper, "Paper Mache Problems in Epistemology: A Defense of Strong Internalism," Synthese, vol. 116 (1998), pp. 27-49. See p. 28.

5. John Pollock, "At the Interface of Philosophy and AI," The Blackwell Guide to Epistemology, ed. John Greco and Ernest Sosa (Maiden, Mass.: Blackwell, 1999) pp. 383-414. The quotation is from p. 394.

6. Ernest Sosa, "Skepticism and the Internal/External Divide," The Blackwell Guide to Episte? mology, pp. 145-157. The quotation is from p. 147. Sosa goes on to describe another version of internalism that highlights accessibility. 7. Laurence BonJour, "The Dialectic of and ," The Blackwell Guide to Epistemology, pp. 117-142. The quotation is from p. 118. INTERNALISMDEFENDED / 15

8. William Alston, "Internalism and Externalism in Epistemology," reprinted inWilliam Aiston, Epistemic Justification: Essays in the Theory of Knowledge (Ithaca, N.Y.: Cornell University Press, 1989), pp. 185-222. See p. 186.

9. Pollock does not make explicit that the internal states to which he refers must be mental states. However, it is reasonable to assume that this is what he has in mind.

10. It has become standard to distinguish between an existing belief (or other attitude) being justified and a person being justified in believing (or having another attitude toward) a proposi? tion whether or not the person actually believes it (or has that attitude). We shall use phrases such as "justified belief to refer to beliefs that are justified and we shall say of a person that he or she is justified in believing a proposition when we mean to say that the latter relation obtains. This distinction will not play a significant role in the discussion that follows. As stated, (S) and (M) are about the justification of existing attitudes. They could easily be reformulated to state internalist constraints on the conditions under which a person is justified in having a particular attitude.

11.Whether (M) implies (S) depends upon details of the supervenience relation which we will not discuss here.

12. Not all philosophers who make this sort of comparison seek an account of internalism with this advantage. James Pryor takes internalism to be accessibilism, notes that internalism in the is a supervenience thesis, and concludes that the two kinds of internalism are dissimilar. See "Epistemology c. 1988-2000," The British Journal for the , (forthcoming).

13. In "Skepticism and the Internal/External Divide," Sosa considers and rejects an argument that has mentalism as a premise and accessibilism as its conclusion. See pp. 146-148.

14. See Richard Feldman and Earl Conee, "Evidentialism," Philosophical Studies vol. 48 (1985), pp. 15-34.

15. In describing and assessing the beliefs in the examples of Section II, we did not say anything about what the individuals had a duty or obligation to believe, what they were permitted to be? lieve, or what they might be praised or blamed for believing. There might be deontological truths of these sorts. We reject arguments for internalism based on the idea that epistemic concepts are to be analyzed in these deontological terms.

16. Other versions of internalism could be considered. For example, some internalists hold that a belief is justified only if the believer is able to formulate good reasons for it.We will not address the merits of these versions of internalism here. We note that a theory requiring for justified belief that the believer not only be able to formulate good reasons but also be able to articulate reasons or an our those persuade others is not internalist theory by standards.

17. "Respondeo Ad Feldman" in Warrant in Contemporary Epistemology: Essays in Honor of Plantinga's Theory of Knowledge, ed. Jon Kvanvig (London: Rowman and Littlefield, 1996), pp. 357-361.

18. "Respondeo," p. 259.

19. "Respondeo," p. 360.

20. Alvin Goldman, "Internalism Exposed," Journal of Philosophy, vol. 96, no. 6 (1999), p. 278. 21. Ibid.

22. It may be that if one were to become conscious of the belief, one would also bring to mind some stored justification that one has for it. Thus, if these stored beliefs were occurrent, they would be justified in the fundamental sense. Whether this justification would happen to accom? pany an occurrent consideration of a belief does not seem crucial. What may be crucial to having 16 / AMERICAN PHILOSOPHICALQUARTERLY

a stored epistemic justification for a stored belief is being capable of recalling a conscious justi? fication, or at least being capable of recalling the key confirming evidence in such a justification.

23. Though it is possible for a stored belief to be justified by one's current evidence, in the usual case, one's evidence for a stored belief will also be stored. It is also possible for an occurrent belief to have only dispositional justification.

24. Richard Feldman, "Having Evidence," in , ed. David Austin (Dordrecht, the Netherlands: Kluwer, 1988), pp. 83-104.

25. Goldman, "Internalism Exposed," pp. 278-279.

26. It is, in the typical case, an internal state that is accessible to the believer, so accessibilist versions of internalism can accept this approach as well. We suspect that many internalists will find the second sort of approach to the problem of stored beliefs more appealing. By limiting evidence to current conscious states, the former view limits severely the number of justified beliefs a person has at any time. We do not regard this limitation as clearly unsatisfactory, given the availability of a dispositional notion of justification to account for the favorable epistemic status of many stored beliefs. We shall continue to present both approaches in the remainder of this paper.

27. For example, they can differ with respect to how readily accessible those states must be. It is also possible to hold that the degree of justification provided by a state is partly determined by how readily accessible it is.

28. Goldman, "Internalism Exposed," p. 279.

29. See, for example, Sosa, "Skepticism and the Internal/External Divide," pp. 145-157. The relevant example appears on pp. 152f. Goldman cites Gilbert Harman, Thomas Senor, and Robert Audi as having raised similar objections. See Gilbert Harman, Change in View (Cambridge: MIT Press, 1986); Thomas Senor, "Internalist Foundationalism and the Justification of Memory Be? lief," Synthese, vol. 94 (1993), pp. 453-476; and Robert Audi, "Memorial Justification," Philosophical Topics, vol. 23 (1995), pp. 31-45.

30. Goldman, "Internalism Exposed," p. 281.

31. The new example resembles one that Ernest Sosa presents involving a generally reasonable a a person who believes conclusion as a result of now-forgotten "tissue of fallacies." Sosa thinks this origin renders the belief unjustified, no matter what the person now thinks about the source of her belief or her general capacities. See Sosa, "Skepticism and the Internal/External Divide," p. 153.

32. Goldman, "Internalism Exposed," pp. 280-281.

33. A third problem Goldman poses for internalism is the problem of concurrent retrieval. It purports to affect only those internalist views that are versions of holistic coherentism. A person has a large set of stored beliefs at any time. Holistic coherentism says that a belief is justified only if it coheres with one's whole corpus of beliefs, including stored beliefs. This leads to a problem for a defender of holistic coherentism who also accepts the deontologically defended claim that one can always find out whether a belief is justified. Ascertaining whether one belief coheres with the rest by bringing them all consciously to mind at once is well beyond the capaci? ties of any person.

This is a problem for holistic coherentism only when it is conjoined with the deontologically defended thesis just mentioned. A holistic coherentist need not accept a deontological conception of epistemic justification, and can simply deny that epistemic status is something that one always INTERNALISMDEFENDED / 17 can find out. The holist can also respond to Goldman's objection by denying that finding out epistemic status so as to comply with any relevant duty requires the simultaneous retrieval of all that the status depends on. Itmight be held to be sufficient for complying with a duty to find out whether belief Bl coheres with one's other beliefs simply to form a true belief, B2, that Bl coheres with the others, as long as B2 itself coheres with the rest of one's beliefs.

In any case, problems peculiar to holistic coherentism cast no doubt on internalism generally. There are, however, related questions concerning the accessibility of stored beliefs that might be raised for other internalist theories, including evidentialism. Here is one of them. Suppose that someone has a conscious belief that is supported by some currently conscious evidence. Suppose further that the person also has a large number of stored beliefs whose conjunction implies the falsity of the conscious belief. This conjunction is too complex for the person to entertain. Under these circumstances, what is the epistemic status of the current belief?

Internalists have ample resources to deal with cases like this, whatever the correct answer to this question is. If beliefs like these are justified, internalists can hold that only currently ac? cessed evidence is relevant to the epistemic status of occurrently believed propositions. So, potentially defeating combinations of beliefs that are not accessed would not undermine justifi? cation. If beliefs like these are not justified, internalists can say instead that all stored beliefs are among the mental items relevant to justification and that any conjunction of them can serve as a defeater of the justification of current beliefs, regardless of whether the individual can consciously consider the conjunction. It is also consistent with M to say that justification supervenes on a restricted class of stored mental items. Perhaps items that are too complex to be retrieved are excluded. In that case, an unbelievably complex conjunction of stored beliefs would not be a defeater. Perhaps only combinations of stored beliefs whose negative relevance to the belief in question has been or could readily be noticed or appreciated count as defeaters. Perhaps, as accessibilists hold, only mental items that are in one way or another accessible can be defeaters. The same variety of claims can be made about what constitutes supporting evidence that one has.

Some of these approaches seem to us to be more promising than others. For present purposes it is not necessary to defend any particular view. We are arguing here for the explanatory power and cred? ibility of internalist theorists. The devil may lurk in the details, for all that we have shown. But in the absence of any good reason to think that internalists must make ad hoc or indefensible claims about stored beliefs, there is no reason to think that there is a general problem here.

34. Alston, "Internalism and Externalism in Epistemology," p. 221. 35. Feldman and Conee, "Evidentialism."

36. See, for example, Laurence BonJour, "Externalist Theories of Empirical Knowledge," Mid? west Studies in Philosophy, vol. 5 (1980), p. 55. In "Epistemology c. 1988-2000," James Pryor calls the view that endorses the higher order requirement "Inferential Internalism" and identifies several of its proponents.

37. Goldman, "Internalism Exposed," Section I.

38. For instance, Michael Bergmann, "A Dilemma for Internalism," APA, Central Division, 2000.

39. Internalists who hold that all evidence is conscious can point to evidence such as the expert's feeling of confidence and sense of familiarity while making the judgment. 40. There is a non-evidentialist view that some internalists find attractive. The idea is that a mental fact about people is that they have fundamental inferential abilities. Perhaps this view could also be described in terms of the ability to see connections. But this view denies that this ability is, or leads to, differences in evidence. This is a mental difference, but not an evidential difference. 18 / AMERICAN PHILOSOPHICALQUARTERLY

These two views can also be applied to the original example about the logic student. We said the TA can see that the original sentence has consequences that the student can't see. This is what accounts for the differences in what's justified for them. As we described the case, we interpreted these facts in an evidentialist way, taking the difference in what they can see as an evidential difference. The non-evidentialist internalist alternative agrees that there is a mental difference between the two, but it characterizes that difference in terms of an inferential skill rather a differ? ence in evidence. It is not essential to a defense of internalism to select between these alternatives.

41. Ernest Sosa, "Beyond Skepticism, to the Best of our Knowledge," Mind, vol. 97 (1988), pp. 153-188. The quotation is from p. 171.

42. The term "recognize" suggests that the classification is accurate. There is no need to insist on an infallible capacity here. If there is some such phenomenon as seemingly recognizing a con? scious quality while misclassifying it, then it is a seeming recognition which supplies the conscious evidence for the classification.

43. An earlier version of this paper was presented at The Creighton Club, where William Alston commented, and to the University of Rochester philosophy department. We are grateful to Alston and to both audiences for their comments. We are also grateful to the Editor and three anonymous referees for the American Philosophical Quarterly for helpful comments.