The enterprise of epistemology is concerned with the norms by which we conduct our intellectual lives. For though the first questions which epistemology asks are about the possibility and nature of knowledge, the notion of knowledge itself has a normative aspect; that is, we are concerned with knowledge because we think it is a good thing to have. And knowledge itself is related to belief; where we are talking about knowledge of propositions, it is widely agreed that knowledge is at least true belief. So the fundamental question of epistemology might be put this way: how should we come to believe what we believe? I shall argue in the first chapter of this book that we necessarily try to believe the truth about each proposition with which we are concerned. Thus, we have a network of terms which it is the project of epistemology to elucidate: `proposition,' `truth,' `belief,' and `knowledge.' This book provides an account of knowledge, and has at least something to say about the other notions as well.


The account of knowledge given here is that knowledge is merely true belief. This account is, I believe, defended here explicitly for the first time. The book as a whole embodies an extended argument for this view. Along the way, it gives a (relatively conventional) theory of belief, re-states what I regard as a traditional answer to scepticism about the external world, and formulates a new view about the nature of human veridical experience. The main thrust of the argument is this: the epistemology of justification - a view which is conveniently encapsulated in the claim that justification is logically necessary for knowledge - is false to our real commitments and practices as doxastic agents.


Though I do attempt to criticize long-dominant trends in epistemology, this essay is, in a certain sense, itself quite traditional. For the most part, I formulate these positions in the terms given within what is termed the analytic tradition, a tradition which, in epistemology and in other areas, I regard as viable. This essay by no means argues for a dismissal or dissolution of the fundamental questions of epistemology: questions about truth, belief, and knowledge. On the contrary, I regard these questions as well-formed and important, and I attempt to address them directly.


In the first chapter, I set out a theory of belief, according to which to believe a proposition is to take that proposition to be true under a representation. This theory, I think, solves some of the outstanding problems surrounding the notion of belief. But the important lessons for the rest of the project are, first, that belief is always verifically oriented, that, as doxastic agents, our goal is always to get the truth. We are often enough deflected from this goal; we may engage in practices of self-deception which allow us to believe, for example, useful falsehoods. Nevertheless, with regard to each of our beliefs, we take it that the believed proposition is true, or else we would cease to believe it. This fact turns out importantly to constrain plausible conceptions of propositional knowledge. Second, in what may seem a trivial point, if belief is a propositional attitude, it always someone's attitude; it is always the attitude of a particular doxastic agent. Beliefs are always things particular persons have, a fact which the epistemology of justification elides at its peril.


The second chapter is intended, among other things, to take on some historical ballast. The epistemology of justification is almost as old as philosophy. And though it is doubtful that any philosopher has explicitly defended the claim that knowledge is merely true belief, many philosophers have mounted compelling attacks on the epistemology of justification. Here I use Descartes (as read by Stroud) as a stalking horse, and emphasize attacks on the method of doubt by Moore, Kierkegaard, Nietzsche, James, and Santayana. This material has the effect of making clear exactly what I mean when I claim that the epistemology of justification is false to our actual commitments and practices with regard to knowledge. One historical locus of the attack on the epistemology of justification has been the defense of the epistemic legitimacy of faith, and it is largely to this material that I appeal in this chapter. In addition, I argue that, whether or not external-world scepticism can be refuted, it need not be.


Then, in the third chapter, I formulate a position that I term "radical externalism with regard to experience." The view, roughly, is that there is no epistemically available inner component of veridical experience which allows us to come to know facts about the external world. I claim, rather, that to experience an external object is to be "fused" with it. (Lest the reader think that I am descending into mysticism, I hasten to add that "fusion" is a technical term I introduce to capture certain aspects of familiar relations.) I argue, with the aid of Austin's attack on Ayer, that certain straightforward claims about the world ("Here's a pig," for example), need not be justified in order to be known.


The final chapter offers a battery of arguments for the claim that knowledge is merely true belief, making use of the conclusions already reached. I am aware that the likely response to the thesis that knowledge is merely true belief is impatience, or perhaps tolerant good humor. But I ask the reader to withold judgement (and, for that matter, good-natured laughter), until these arguments have been canvassed. You may, when all is said and done, find the thesis as unconvincing as you expect. On the other hand, I hope and believe that you will find arguments here that seriously challenge your assumptions.


Finally, I want to emphasize again (as I will be emphasizing relentlessly throughout) that I regard the fundamental questions of epistemology as being "What ought we to believe?" and "How ought we to come to believe it?" And I want to emphasize a neglected aspect of these issues: the question of who "we" are. Who is it whose intellectual lives epistemology seeks to give norms? The answer I give to this is that we are particular human beings. This may strike some as trivial or obvious, but I will argue that ignoring it has led to fundamental distortions of the epistemological enterprise. The persons to whom epistemology addresses itself are real as opposed to hypothetical persons, flesh and blood individuals as opposed to theoretical contructs. Now I am not going to be directly concerned here with what a human being is. I am not going to try to answer, for example, the question of whether human beings are merely configurations of material, or rather of ectoplasm. But I am a particular human being - "merely" physical or not - and I address particular human beings.


So the problems of epistemology are practical problems. Epistemology asks, what should we, you and I, believe, and how should we come to believe it? It asks, what, if anything do we know, and how do we know it? After all, in the best case we would like to act on knowledge, or at least on reasonable belief, rather than on mere speculation or guesswork. The epistemological project then, should help us see how we ought to live, not in the moral sense of how we ought to act, but in the sense of on what we ought to act, morally or otherwise. Epistemology thus explicates the material for deliberation and for action.


I gratefully acknowledge the assistance of Jeffrey Tlumak, Joshua Tonkel, Clem Dore, John Post, and Michael McKenna, who commented on various versions of the material presented here. Jim Montmarquet read various versions of various chapters, and his comments were invaluable. The original impetus for this book came from seminars with Renford Bambrough and Jim Cargile when I was a graduate student at the University of Virginia. Many ideas presented here germinated in various discussions and classes with Richard Rorty, my dissertation supervisor. I think it is only fair to acknowledge that none of these people agree with my claim that knowledge is merely true belief. I am especially grateful to the Vanderbilt University philosophy department, including John Compton, John Lachs, Idit Dobbs-Weinstein, Don Sherburne, and Charles Scott for providing the atmosphere in which these ideas could be developed. In addition, I would like to thank my wife Rachael and my daughter Emma, who not only tolerated my work on this manuscript, but put up with innumerable questions about what they knew and how they knew it.





Chapter I


Belief and Truth



I am going to argue in this chapter that belief is conceptually connected to truth. Believing a proposition p and taking p to be true seem to me to amount roughly to the same thing, and I hope to use this insight, rudimentary though it is, to generate something like an analysis of belief as a propositional attitude. Though I will explain this analysis, I will not provide a full-dress explication and defense. That is because the purpose of discussing belief here is simply to help us understand the concept of knowledge. Thus, I will emphasize the elements of this theory of belief that are relevant to that project, and omit various problems that might arise in other respects.


Among the projects I will not take up in this chapter is the project of establishing that belief is a propositional attitude, much less will I try to establish that there such a thing as belief, or that talk of belief is a (scientifically or philosophically) legitimate idiom. That is, I am not going to grapple in any elaborate way with the arguments of Stich, Dennett, Churchland or other "doxastophobes" (to borrow a phrase from William Lycan). It seems to me that belief is a perfectly good rough and ready notion, and that it is neither possible nor desirable to do without it. It may well be possible to give a physiological or functional characterization of what we now term belief-states that would eventually let us get away without using the term `belief' in scientific contexts. It might turn out that, when such a characterization is achieved, we might see that at the functional or physiological level, there is nothing that all such states have in common. And there are indeed serious problems with the treatment of belief as propositional attitude, though that is the sort of account I endorse. But that does not show that we shouldn't use the notion of belief in scientific contexts, much less in everyday conversation, any more than the fact that tables and chairs are made of quarks shows that we ought to stop talking about furniture. Again, I mean to be concerned with practical matters about how we conduct our intellectual lives; such a project cannot begin by accepting a radical shift in the normal ways in which we talk about those lives. Quine's dictum that "philosophical clarity is . . . largely a result of avoiding mental entities," has long played a part in how philosophers treat questions in epistemology.(1) But I suggest that this chauvinism against the mental arises out of prejudice. And if we go on talking in mentalese, that does not mean that we ignore conceptual difficulties, that we do not await the promised reduction, or that we believe in poltergeists.








I. The Attitude of Belief


So I take belief to be a propositional attitude. Propositions I will treat roughly as whatever it is that bears truth value. So belief is a relation between a cognitive agent and a bearer of truth value. I will think of propositions as being modally individuated: two sentences express the same proposition if and only if their truth value is the same in every possible world. Now there are fairly severe problems with this, not the least of which is that, according to what I have said so far, every necessary truth expresses the same proposition. For example, the proposition expressed by the sentence `2+2=4' would seem to be the same as that expressed by the sentence `3+3=6,' since both of these are true in every possible world. But it seems that it is one thing to believe the one and another thing to believe the other. In addition, it seems that all logically equivalent sentences will express the same proposition. But this seems obviously false if propositions are taken as the objects of belief, since one can clearly believe believe p but not believe q even where p is logically equivalent to q. There are various ways to finesse such problems.(2) And there are various ways to try to individuate propositions in a more fine-grained way (for example, the "structured proposition" account put forward by Mark Richard, Scott Soames, and others).(3) I think the gist of my account will be basically unaffected by these debates, or at least I think the account could be refined to take account of a good theory of how propositions ought to be individuated. If the reader is uncomfortable with an ontology that includes entities such as modally individuated propositions, I invite him to consider propositions as theoretical, explanatory entities.


I am going to start with the notion that a proposition may be taken in one way or another. In this sense, one may take a proposition to be true or false, one may take it to pick out a desirable or undesirable state of affairs, one may take it as funny, mind-boggling, and so forth. Now to the extent that this notion is taken as primitive, or, to put it in less sanguine terms, to the extent it remains unexplicated, the theory of belief I give will be in a certain sense stunted. But it is, to begin with, obvious that "taking" a proposition in one way or another is not simply identical to believing it (since there are many ways of taking a proposition), so the account is at least not baldly circular. And I do want to say something by way of putting some meat on the bones of this notion.


To take a proposition in one way or another is to have an emotional response to it, to entertain it and respond to it emotionally. For example, to desire that a certain state of affairs obtain is (at least in some cases in which one's desire is conscious) to have a certain emotional reaction to entertaining a proposition, to harbor a certain attitude toward that proposition. An account of the affective content of this particular propositional attitude would be an account of desire. One way of putting the point is to say that to take a proposition in one way or another is to have a certain pre- or extra-cognitive response to it. (Ultimately, I would try to impugn the distinction between the affective and the cognitive. I use it here to convey a sense of what I am after.) In the case of belief, the response involved is a certain sort of commitment. We are all familiar with various sorts of emotional commitment, or at least with emotional commitments to various objects: countries, careers, people. Taking a proposition to be true is that kind of thing, with a propositional object. I am not going to offer any elaborate argument to the effect that belief is a matter of emotional commitment (though I will have much more to say on the topic in the next chapter). But I will remark that, if belief were not a matter of emotional commitment, people would be more rational than they are. For reasons that will emerge throughout this book, I do not think that would necessarily be desirable.


With this is in mind, let us consider an obvious way of connecting belief and truth by means of a definition of belief:



(A) S believes that p if and only if S takes p to be true.



This, however, will not do, as John Perry and many others have given us reason to see. Perry says that he once followed a trail of sugar around a grocery store trying to find the shopper with the leaking bag. After going several times around the same counter and watching the trail grow thicker, he realized that he was the shopper he was trying to catch. He believed that the shopper with the torn bag was making a mess, and though he was all along that very shopper, he only later came to believe that he himself was making a mess. It is impossible to explain why he suddenly came to reach down and adjust the bag in his cart unless we understand how his belief shifted. However, did the propositions he believed shift? Well, as a matter of fact I think in this case they did. There is a possible world in which Perry was following someone else, that is, in which the description `the shopper with the torn bag' picks out a different individual. But consider the case where he comes to believe that John Perry is making a mess. This explains his behavior, he points out, only on the supposition that he believes he is John Perry (and not, for example, Charles de Gaulle). So if the indexical and the name are "rigid" (if they pick out the same individual in all possible worlds in which that individual exists) then there is no shift in the list of propositions Perry believes. But there is a shift of belief.(4) At a minimum, this shows that (A) does not capture the ordinary sense of `belief' in some cases where we think that a person's beliefs can be employed in an explanation of their actions. This is because, though Perry does not seem to come to believe any new propositions, he seems to undergo a shift of belief that explains his change of behavior.


It is possible also to make the point without the use only of names and indexicals. Consider Nelson Goodman's claim that the identity of a literary work consists in the way that it is spelled.(5) That is, two items are both copies of the same work just if they are spelled the same, if they are notationally indiscernible. Now consider these two sentences: `Scott is the author of Waverly' and `Scott is the author of: . . . ', where the ellipsis in the second sentence is filled in with a complete, correct spelling of the novel Waverly. If Goodman is right about literary works and I am right about propositions (both admittedly shaky assumptions), these two sentences express the same proposition. But it is perfectly possible to believe one and not the other, if one is laboring under a misapprehension as to the contents of Waverly.


For reasons such as these, Perry distinguishes between the belief states of a person and the propositions she believes. That is surely correct. I will explicate this distinction by saying that S's belief-state consists in S taking some representation of p to express a true proposition. And now we can think of the notion of taking a proposition in one way or another as being concerned with responses to representations of propositions, to propositions under representations. This is a good thing, since propositions, if they exist at all, are some sort of abstract entity. If they are construed, for example, as sets of possible worlds, then it ought to be obvious that such things cannot appear in anyone's mental economy. But representations of propositions, for example, tokens of sentences expressing propositions in some language, can so appear. (At least, it is plausible to hold that they can. I do not intend to address various problems about this.)


I will use single quotes to refer to representations of p: `p' is a representation of p. Typically, a representation of p will consist of some mental or physical token of a sentence that expresses p relative to the conventions of the language of the sentence. (I am, to repeat, not going to take up various quandaries about "mental tokens.") However, there is no reason to limit representations of propositions to sentences of public languages. For example, if there is a "language of thought" a la Fodor, presumably it must be capable of representing propositions (if there are propositions). Furthermore, perhaps mental images can be employed as representations or segments of representations of propositions. And on some accounts, objects in the world can be part of propositions, and, perhaps, they can also be parts of representations of propositions.(6)


Again, we need to introduce representations of propositions as well as propositions themselves here for notorious reasons having to do with the substitution salva veritate of co-referring expressions in belief contexts. If I take `a is F' to express a truth, I do not necessarily take `b is F' to express a truth, even where a and b are identical. For example, if I take `Groucho wears a greasepaint mustache' to express a truth, I do not necessarily take `Julius wears a greasepaint mustache' to express a truth, even though Groucho and Julius are identical. This can be seen from the fact that if I do not believe Groucho and Julius to be identical, I may sincerely deny that Julius wears a greasepaint mustache, and it seems implausible to claim that I believe what I sincerely deny. And if `Groucho wears a greasepaint mustache' and `Julius wears a greasepaint mustache' express the same proposition, it would follow that it is possible to believe and disbelieve the same proposition at the same time under different representations. The same result follows on most version of the "structured proposition" account, because although on most such accounts the structure of the proposition mirrors the structure of its representations, the structure of the proposition is not sensitive to the the spelling of proper names. Let us, then, build these results into a new proposal for a definition:



(B) S believes that p if and only if S takes some representation of p to express a true proposition.



It is worth noting that though I formulate this definition largely in response to Perry's "essential identical" problem, it certainly does not give a solution to the fundamental difficulty Perry raises. That is, it does not explain why indexical expressions such as `I' are ineliminable from the propositional attitude contexts in which they occur (or at least, are sometimes ineliminable). Another way to put this is that it does not account for the peculiarities of function that characterize representations of propositions that include tokens of indexical expressions. But I do not see any reason to suppose that different representations of the same propositions cannot play dramatically different roles in one's mental activity or behavioral outputs. The point is, I think, intuitive. For example, suppose that I believe certain propositions which I represent with (1) `Groucho wears a greasepaint mustache' and (2) `Groucho says "I would have thrashed him to within an inch of his life, but I didn't have a tape measure."' Now it is very likely that I also believe a proposition that I might represent as (3) `Groucho wears a greasepaint mustache and says "I would have thrashed him to within an inch of his life, but I didn't have a tape measure."' But if instead of under (2), I believe the same proposition under the representation `Julius says "I would have thrashed him to within an inch of his life, but I didn't have a tape measure"' then it would not be surprising to find that I did not believe the proposition represented by (3) under that representation. And it very well might be that we would notice aspects of my movie-renting behavior that reflected these differences. So I do not think that the problem of the essential indexical is different in kind than other problems which arise here, and which (B) allows us to deal with in a fairly perspicuous way.


But (B) still does not provide a sufficient condition for belief. To see this, consider the following case. I look at a representation of the proposition that snow is white (for example, I look at a token of a sentence `snow is white' on a printed page). However, I think it represents the proposition that the moon is a balloon, which in turn I take to be true. In this case, I satisfy (B) with regard to the proposition that snow is white, but I do not believe it in virtue of satisfying (B).


This is because the token of `snow is white' does not function for me as a representation of the proposition that snow is white, though it indeed represents that proposition.(7) In order for it to so function, I must take the term `snow' in the representation to refer to snow, and the predicate `white' in the representation to refer to whiteness. If we accept a causal theory of reference, we will say say that I must, in order to believe that snow is white by taking some `snow is white' to express a true proposition, be connected by the right sort of causal chain, through speakers of the public language, ultimately to snow. (It is no part of my brief to set out or defend such a view here.) The referring expressions of the representation as used by me must be secured the proper referents in order for that representation to function for me as a representation of the proposition in question.


If the definition is to provide a sufficient condition for belief, then, the representation must function for the believer as a representation of the propositional candidate for belief. Thus:





(C) S believes that p if and only if S takes some item which functions for S as a representation of p to express a true proposition.



Now this introduces an element of vagueness into belief. For there are various ways in which and various degrees to which some item can function for me as a representation of a proposition. (C) entails that I have some semantic mastery over the components of the item which functions for me as a representation of p. I cannot believe that 2+2=4 by believing that `2+2=4' expresses a true proposition if I do not have some mastery of these symbols. However, it turns out to be very hard to say exactly what degree of mastery is required. For example, I think to some degree it is true of me that I believe that the theory of relativity is correct. On the other hand, I can say almost nothing interesting about the theory. I believe it in an attenuated or impoverished sense. As I find out more and more about the theory, as I acquire more and more mastery over the terms in which the theory is couched (in part, by coming to believe other propositions in representations of which those terms appear), my belief that the theory is true comes to be more and more full-fledged.


Now this suggests a problem for the propositional attitude account of belief that has been raised by Dennett and others. For it seems that on it, I do not know what propositions I do believe.(8) If we are pressed to postulate belief states or representations of propositions as well propositions themselves, then it seems that our access to the latter is always mediated by the former, and it becomes impossible to know what we in fact believe. This sounds like it is going to lead to some very disconcerting brand of scepticism.


But the problem here is, I believe, spurious. In fact, one can find out more and more about what propositions one does believe. Again, if I was concerned to do so, I could find out more about the theory of relativity. This would be, among other things, to increase my access to the propositions I already believe by, first of all, multiplying my representational abilities with regard to those propositions. That is, one thing that happens in the process of learning about the theory of relativity, in making my belief that it is true more and more full-fledged, is learning a vocabulary, learning, for example, which representations express the same proposition. Secondly, learning about the theory allows me to represent related propositions that were previously outside my repetoire. It may be that I never become an infallible reporter of what propositions I do believe, but I can clearly become more and more reliable in this regard.


One might think of this as analogous to coming to know what some historical figure looked like by examining portraits of him. This is surely possible; that is, it is surely possible in general to come to know something about x through the examination of representations of x.(9) Furthermore, we can have access (unmediated, I think, by representations, though I reserve this question for chapter three) to the states of affairs in the world to which propositions correspond or fail to correspond. Where we have such access and take the trouble to exploit it, our beliefs can become very full-fledged indeed, and we may become extremely reliable reporters of what propositions we do believe. I take this to mean, for example, that if we listed our beliefs as a set of sentence tokens, it would be rare for one such token to express a proposition and another to express the negation of that proposition. By critical examination of our beliefs and by empirical investigation, we can render our beliefs more and more consistent. In fact, this is part of what we mean by rationality.


(C) is, I think, very nearly right. Nevertheless, we may want to add one more qualification. If I take some representation of a proposition to express a truth due to, say, a momentary lapse in alertness, we do not necessarily count that proposition as being among my beliefs. For example, suppose I am listening to a speaker that I admire and am swept up by her rhetoric into a momentary agreement with some strikingly idiosyncratic claim. On reflection, I quickly realize that the claim is profoundly incompatible with some of my deeply held beliefs. It might be appropriate in such a case to say that I didn't believe the initial claim at all. Of course I am by no means denying that people have profoundly dissonant beliefs: I mean only to deny that beliefs can be more or less wholly stray or casual. A belief is constituted as a belief within a system of belief. This does not necessarily amount to logical consistency, much less explanatory coherence, but it is nevertheless worth noting.


So here is what I take belief to be:



(D) S believes that p if and only if there is some item which functions for S as a representation of p and which S reflectively takes to express a true proposition.







II. Some Clarifications


I do not take (D) to entail that S has any beliefs about propositions per se; that is, S does not have to be aware that the process she is undergoing consists of taking representations to express true propositions. All that is required is that her mental state is properly characterized in that way. Furthermore, S does not have to have any theory of truth. The `takes' operator introduces an opaque context; however S thinks about truth, that is the way she thinks about the propositions she in fact believes.


This characterization of belief treats believing as an inner act analagous to assertion. That is, I think of taking a `p' to express a truth as analogous to asserting a `p' inwardly. Alternately, the belief state may rely on an external representation of the believed proposition, as when I take some sentence-token in the morning paper to represent a true proposition. Here, we might think of the act of believing as analogous to assent.


But in both cases, believing is an inner event. We may well come to know what people believe by their outward behavior, including linguistic behavior. If you tell me that you believe that it's going to snow tomorrow, or if you assent to my claim that it will snow tomorrow, or if I see you stockpiling chicken soup and toilet paper, I may know in each case perfectly well what you believe. But your belief itself does not consist of your behavior, and your dispositions to behave are not part of your belief per se.


The sort of extreme behaviorism I am rejecting here was at one time fashionable (and has occasionally been revived(10)), but it is highly implausible. R.B. Braithwaite, for example, wrote `I believe that p' means "(1) I entertain p" and "(2) I have a disposition to act as if p were true."(11) This just cannot be right. Consider an experiment in which I cage a human subject and deprive him of food for several days. Then I start to feed him whenever he assents (or leaps up in down with glee, or whatever) when presented with `2+2=5,' or counts five when presented with two blocks, and then another two blocks, and so forth. If I do this long enough, the subject will acquire a disposition to behave as though 2+2=5. But that hardly shows that he believes it (even if he does entertain it).


Another popular technique for trying to cash out belief in behavioral terms is the counterfactual bet: what you'd bet on a proposition shows how committed you are to it. However, people are notoriously prone to venture sums larger than they can afford on a mere throw of the dice without any confidence whatever that they will win. Here we might retreat to a notion of rational betting behavior. But this would beg certain questions I want to keep open, and it would represent a retreat from the question of how we actually conduct our doxastic lives to a question about how we ought to conduct them. That is, it would explicate not the notion of belief, but the notion of knowledge; it would tell us what we ought to believe, and leave our actual beliefs unaccounted for.


Consider a proposal, relevantly similar to mine, from Nathan Salmon. He writes: "To believe a proposition p is to adopt an appropriate favourable attitude toward p while taking p in some relevant way."(12) Now this is good as far as it goes. But I think that my version has the merit of specifying which favourable attitude is appropriate (taking to be true) and of specifying which way is relevant (via a representation). It is important once again to notice that belief is always tied to truth. I may approve of the proposition that I look like Mel Gibson and think like Spinoza, but no matter how it warms my cockles, I do not believe it. As Dorothy Parker wrote: "Love is a thing that can never go wrong, and I am the Queen of Rumania." The only dimension of approval that is relevant to whether one believes is the the verific dimension. Of course, I am even now in the process of systematically deluding myself into believing that I look like Mel Gibson and think like Spinoza. But I will only believe it when I take it to be true.



III. David Austin's Objection


Now one fairly obvious sort of objection to this account, and particularly to the claim that belief-states are to be characterized by operations over representations of propositions, is that it leads directly to contradictions. For as I have admitted, indeed emphasized, it seems possible both to believe and not to believe the same proposition at the same time. This may appear to be a reductio ad absurdum of (D).


There are at least two ways in which the claim that someone believes and does not believe the same proposition at the same time can be construed. First of all, one might be held both to believe and to disbelieve the same proposition, that is:



(1) Bxp & Bx~p



where `B' expresses the belief relation `x' picks out a doxastic agent, and `p' a proposition. Now (1) is not itself a contradiction. But it certainly expresses an undesirable state of affairs. Insofar as x is rational, she will seek to drop either the belief that p or the belief that ~p. We are in a position now to see very clearly why this should be the case. To believe a proposition is to take that proposition to be true. But if she is rational, the agent will become aware, as she becomes aware that she believes both p and ~p, that she has a false belief (though she may not be in a position to decide which belief is false, in which case the rational procedure is to suspend belief with regard both to p and its negation). For example, if I became aware that I believed both that God existed and that God did not exist, I would probably start trying to evaluate the arguments on both sides.


Now a much more disconcerting case seems to be



(2) Bxp & ~Bxp



(2) expresses a direct contradiction; it simply cannot be true. And it might be held that (2) is entailed by (1). More precisely, it might be held that Bx~p entails ~Bxp. This inference is pretty clearly fallacious. But let us examine a specific argument that attempts to show that the propositional attitude account of belief leads to contradictions.


David Austin gives the following example, which he calls the "two tubes" case.(13) Smith is the subject of an experiment designed to test his ability to focus each of his eyes independently. He is sitting in front of an apparatus which consists of a screen, from the back of which emerge two tubes. Each of his eyes looks through one of the tubes. He is supposed to look simultaneously through both tubes and report what he sees with each eye. As it happens, the two tubes are oriented so that he is looking at the very same red spot through both tubes, though he has no way to be aware of this fact. He names the spot he sees through his right eye `this' and the one he sees through his left eye `that.' He believes



(P1) This is red and that is red.



Futhermore, as Austin puts it, Smith is somewhat pedantic, and he also believes that



(P2) This is this and that is that.



Now consider the claim that the demonstratives `this' and `that' are rigid designators. (Again, I will not argue for this, but I, like Austin, regard it as plausible.) Then it follows that the representation `this is this' and the representation `this is that' express the very same proposition. But it is not the case, it seems, that Smith believes that this is that. In fact, he explicitly denies that he believes it; he has no reason to believe it. That is, Smith appears to satisfy (2) with regard to the proposition that this is that; it is both the case and not the case that he believes it. (This statement of the two tubes case is simplified from Austin's version, but it captures the thrust of the example construed as an objection to the present account of belief.)


Now in fact, (D) can account for this apparent contradiction rather smoothly. For notice, (2) construes belief as a dyadic relation between a person and a proposition. But (D) in fact gives an account of belief as a triadic relation between an epistemic agaent, a proposition, and a representation of a proposition. `This is this' and `this is that' in fact represent the very same proposition. But Smith believes that proposition under the representation `this is this' and fails to believe it under the representation `this is that.' So he instantiates this schema:



(3) Bxpr & ~Bxps



where `r' is a constant picking out `this is this' and `s' is a constant picking out `this is that.' This is no contradiction. I conclude, therefore, that the accusation that (D) leads directly to contradictions is misplaced. I will refer to cases both of (1) and of (3) (where we move back to variables and close the formulae with quantifiers) as cases a doxastic dissonance, and I will discuss such cases in the next chapter.


IV. Tacit and Occurrent Beliefs


Beliefs are customarily divided into tacit and occurent. Roughly, my occurrent beliefs are those which I am right now or have been quite recently in the process of mentally traversing, or muttering to myself, or endorsing, and so forth. Tacit beliefs, on the other hand, are those which under certain circumstances I would traverse, mutter etc. For example, it may never occur to me that there are more than seven people living in Ulan Bator, but a skillful questioner could get it out of me that I do believe that.


Now there seems to be a rather yawning gap between these two senses of belief. I seem to have pathetically few occurrent beliefs, limited to the number of things I can mutter approvingly to myself simultaneously. Perhaps in fact I have at most one occurrent belief at any given time, because of limitations on my mental voice, or my mental ear. On the other hand, I clearly have an infinite number of tacit beliefs: I believe that more than eight people live in Ulan Bator, more than nine, and so forth up to a couple hundred thousand or so, and similarly for every major metropolitan area of which I've heard. For any whole number equal to or greater than three, I am disposed to believe that it is greater than two. Neither notion seems to capture the usual intuitive notion of belief. For example, though I believe that Wittgenstein was an obscurantist, this is not a mere disposition, like my belief that there are more than one hundred twenty one people in Ulan Bator, but it is not, by the same token, a constant occurrence (I don't (always) go around muttering "Wittgenstein was an obscurantist" to myself).


So it seems worthwhile trying to formulate an intermediate sense of belief that captures the way I believe that Wittgenstein was an obscurantist. I suggest, therefore, the following division:





S believes occurrently at t that p if and only if there is some item which functions for S as a representation of p which S takes at t to express a true proposition.


S believes tacitly at t that p if and only if there is some item which functions for S as a representation of p which S would take to express a true proposition if it occurred to him at t.


S believes enduringly at t that p if and only if there is some item which functions for S as a representation of p such that S takes that item at t-1 (that is, some time previous to t) to express a true proposition and which S would take to express a true proposition if it occurred to him at t.



So I characterize tacit beliefs counter-factually as what one would believe occurrently under certain circumstances. This relieves us of the problem that, if tacit beliefs are counted among our mental states, we would seem to have an infinite number of mental states. For though I have each of the dispositions I have in virtue of some state of me, it does not follow that there must be a distinct such state for each distinct disposition. My enduring beliefs are then characterized as what I have believed occurrently and would believe occurrently again if they occurred to me. My belief that Wittgenstein is an obscurantist is of this sort. I take what I term "enduring" belief to express the primary notion of belief. Except in rather special circumstances, where I ascribe a belief to someone, I ascribe an enduring belief.



V. Acceptance and Self-Deception


The connection of belief to truth for which I have been arguing is established by a very rudimentary fact: the fact that to say "I believe p, but p is false," is to contradict oneself. To say that is not to utter a contradiction, since both of the conjuncts can be true. But it is to affirm what one denies, which is what I mean by "contradicting oneself." Now of course it would be silly to deny that people often believe things they want to believe for no good reasons. In fact, that seems to me to be a typical situation. We are not truth machines but extraordinarily complicated organisms, and we rarely even have access to the sub-conscious processes that produce our beliefs. Again, belief is an expression of passion, and much of our emotional lives is hidden from us. If asked why he believes that God exists, for example, a theist may come up with a variety of arguments, but these very likely have nothing at all to do with the causes and motivations of his belief. The cause is perhaps found in his early religious training; the motivation is perhaps found in his inability to accept a world without an overall purpose or to accept his own mortality. But one thing is absolutely certain: he does not both believe it and take it to be false under the same representation at the same time. It seems to me that we rarely function as impartial seekers after truth. We often believe what we need to believe for no reason worthy of the name. Nevertheless what we believe is precisely what we take to be true, and vice versa.


We need, then, to introduce a familiar distinction between the causes of our coming to believe something and the reasons for which we believe it. The reasons for which one believes p are always taken by one to tend to establish the truth of p. The causes of one's coming to believe that p, however, may well have nothing to do with truth. I often enough believe things because I want or need to believe them, but never for the reason that I want or need to believe them.


This raises the issue of self-deception. There are indeed ways that I can start to get myself to believe that I look like Mel Gibson and think like Spinoza. But believing that proposition is not possible without a process by which I come to persuade myself of its truth. Again, it is incoherent for me to say that I believe it, though of course it's false. The reasons I have for believing it may be extraordinarily bad ones, of course, but I believe only when I take to be true. Consider the following example, which has been discussed by Jack W. Meiland and by John Heil. Heil frames the case like this:



Sally and Bert have been happily married for fifteen years and have every reason to look forward to continuing connubial bliss. One day, however, Sally notices a long blonde hair on Bert's coat. On a later occasion she discovers in the same coat a lipstick-stained hankerchief and a matchbook from an intimate French restaurant that she has never visited. These, together with certain other bits of evidence, seem plainly to warrant the conclusion that Bert has been seeing another woman.


Sally, however, while appreciating the import of the evidence against Bert, believes passionately in her marriage. It is, she thinks, worth preserving even if the cost is high. Further, Sally recognizes that she is the sort of person who cannot easily conceal suspicions. If she were genuinely to believe that Bert might be unfaithful, she would come to treat him with coolness and indifference, and the marriage, already teetering, would not survive. Because this prospect is unacceptable to Sally, she decides to believe, despite a mounting tide of evidence to the contrary, that Bert is not seeing another woman.(14)



Now, as Heil immediately points out and as is notorious, the notion of "deciding to believe" is problematic. As a matter of fact, one cannot by a sheer act of will come to believe some proposition that one regards as false. That is, strong forms of what is known as "doxastic voluntarism" are implausible. I will discuss this issue in somewhat more detail in chapter four. But for now we can note that one reason that strong doxastic voluntarism is implausible is, as we are now in a position to appreciate, that the notion of believing p entails taking p to be true. There is no such thing as non-verific acceptance or belief. Sally cannot do what Heil describes her as doing without launching into a process of self-deception, that is, without systematically ignoring reasons to regard it as true that Bert is having an affair, and systematically cultivating reasons to regard it as false that he is. She might, for example, resolve never to look in Bert's pockets, or attempt to persuade herself that blonde hairs have a variety of sources. Her reasons may be extraordinarily bad. But the point is, she is going to have to persuade herself that it is false that Bert is having an affair; when she has done so, and only when she has, will she stop believing it. She may end up acting as though (or trying to act as though) she believes Bert to be faithful, but she is powerless to believe it except as the result of a process of self-deception. Acting as though something is true has nothing to do with accepting or believing it, I would argue (and have argued). Belief is always tied to truth.





VI. The Inductive Argument and Subjective Probability


Once we have distinguished the causes of belief from reasons to believe, one sort of objection to the present account falls away. But there is another, and in some ways much more compelling objection to the claim that belief is conceptually tied to truth, which might be termed the inductive argument.(15) It goes like this: I have often enough come to realize that some belief of mine is false. Therefore, if I am reflective, I will tend to doubt that all the beliefs I currently hold are true. (That is, I may come to believe that it is not the case that all of my beliefs are true.) And if I follow out the implications of that claim, I may come to doubt of each of my current beliefs (except perhaps some very fundamental ones) that it is true. To frame the argument in the most compelling way, let us say that I accept a frequency theory of probability. And let us say that I have an infallible technique for assessing the truth or falsity of beliefs up to some time t in the past. And let us postulate further that in making that assessment I come to see that up until t my beliefs were pathetically inaccurate, that ninety percent of them were false. I infer that there is a ninety percent chance that any given belief I currently hold is false. But I continue to believe the propositions I did at the beginning of the exercise.


Now first of all, I think it is clear that I would be perverse to continue to believe what I believed at the beginning of the exercise. In fact, the inductive process described is close to the traditional Pyrrhonian methodology for producing absolute suspension of belief. It could be employed to motivate a withdrawal of emotional commitment from more or less the entire structure of one's beliefs. (Though, as I will argue in the next chapter, that is a prodigiously difficult project.) It seems likely that if I were to sincerely carry out such an exercise, I would begin to doubt my beliefs in a wholesale manner. Nevertheless, I agree that it is possible to work oneself into a state in which one comes to think that there is a ninety percent chance that each of one's beliefs is false (excluding the belief that there is a ninety percent chance that each is false) while nevertheless retaining those beliefs. That is, it is possible to believe p under some representation while also believing under that representaion that there is a ninety percent chance that p is false, though such a situation is no doubt unusual. However, this provides no reason to divorce belief from a subjective assessment of truth, but only from a subjective assessment of probability.


Richard Swinburne, for example, argues that one believes p if and only if one believes that p is more probable than its alternatives, where the alternatives generally include not-p.(16) And Jonathan Bennett writes that "I am using `believe that p' to mean `regard p as highly likely'."(17) One apparent advantage of views that connect belief to subjective probability is that they makes sense of degrees of belief, whereas my formulation makes it appear that belief does not admit of degrees (I shall return to this point momentarily). Now Swinburne's formulation faces difficulties, I think. We have seen that it is possible to believe that not-p is more probable than p and still to believe p. For example, suppose again that I am an afficianado of frequency theories of probability and I have observed that seventy-five per cent of the balls drawn from a barrel of black and white balls have been black. Then I might believe that there is a .75 probability that the next ball will be black. But I might have a sudden irresistible premonition that it will be white.


Swinburne and Bennett might argue that, if in fact I do believe that it will be white, if I'm suddenly (though inexplicably) certain that it will be, then I believe that the probability that it will be is 1, or nearly so. But first of all, I might just deny this. (It's possible to have irrational beliefs.) And second, this would be to collapse Swinburne's view into something like my own, since it sounds as though believing is now simply tantamount to `taking to be true'. If the probability of p is 1, p is true. Furthermore, it is perfectly possible for me to take p to be (marginally say) more probable than not-p and not to believe that p. I may simply be lazy about following out my own probability judgements, or I may have little confidence in those judgements.


But this is not, I think, the case with truth. It is sheer incoherence to affirm that although I reflectively take it that `p' (which functions for me as a representation of p) expresses a truth, I don't believe the proposition it represents. At that point I'm merely gibbering, speaking nonsense. Either I take `p' to express a truth or I don't. To believe p entails that I take some representation of it to express a truth. (Note that p can very well be a proposition about probability; I can believe that q is highly probable without believing that q. That is, I can take some `q is highly probable' to express a truth without taking any `q' to express a truth.) Now I may also take p to be false if I take some other representation of p to be false. But in such a case, if I am rational, I cannot be aware that the two representations represent the same proposition.


To relate my proposal to Swinburne's, it seems to me that while there may be a defensible notion of subjective probability of the kind to which he appeals, its only relevance to belief is that I believe p if the subjective probability I assign to p through a representation of it crosses the threshold at which I am willing to call it true. Up to that point, subjective probabilities are irrelevant to belief. (And in fact, if I am sufficiently perverse, there is no such point short of 1.) We are aided here by the fact that, if in fact there is an introspectively viable notion of subjective probability, it is much less fine-grained in the usual case than various objective measures of probability. It consists, as Bennett's treatment suggests, of categories such as "likely" and "very likely." We're rarely in the business of assigning .57 to p and trying to figure out whether we believe it.



VII. Degrees of Belief


Again, however, it is an advantage that Swinburne can capture the notion of degrees of belief, whereas it seems that on my view that notion makes no sense. Nevertheless, there are at least two ways that the current view can make sense of degrees of belief. First, it might be suggested that, whatever subjective probability I regard as sufficient for me to take `p' to express a truth (e.g., "very likely"), any subjective probability higher than that ("extremely and emphatically likely") increases the degree of my belief. But again, this leaves out of account the fact that I can believe something, say, on a hunch, while simultaneously holding it to be improbable. I may agree with the oddsmakers that Pretty Baby is a longshot in the seventh race. But I may still believe that Pretty Baby will win. Thus, I come to the second suggestion. We might understand degrees of belief in terms of resistance to revision; the less willing I am to revise my belief, the more dear it is to me, the greater the degree to which I believe it. Resistance to revision is connected in some cases to subjective probability, but it need not always be. For example, I may have an extremely high subjective probability for the proposition that my wife won't leave me, but this may be quickly revised when I get home and find the furniture gone. In contrast, I might assign a fairly low probability (though well over .5) to the claim that the universe is a naturalistic system, but this belief may be practically unrevisable in the face of any particular experience. For example, it might easily withstand Kweskin bending keys or Shirley Maclaine speaking in voices from beyond the grave.


However, there are obvious difficulties with this last suggestion; I, for one, do not know how to specify exactly to what extent any given belief is resistant to revision. So I arrive at a final move: I will simply bite the bullet. There are no degrees of belief. This view was famously defended by Newman in An Essay in Aid of a Grammar of Assent. (What Newman terms `assent,' I call `belief.' `Assent,' however, is felicitous; it relentlessly emphasizes the verific orientation of belief.) He writes: "We might as well talk of degrees of truth as of degrees of assent."(18) And in a passage that anticipates several of the dominant themes of this book, he writes: "if assent is acceptance of truth, and truth is the proper object of the intellect, and no one can hold conditionally what he holds to be true, here too is a reason for saying that assent is an adhesion without reserve or doubt to the proposition which is given" (114).


Now this seems to contradict the obvious fact that there are degrees of commitment involved in belief. But Newman suggests an admirably clear solution to this problem, one that I have already mentioned in passing. To "partly" or "conditionally" or "to some extent" believe that p is not in fact to believe p at all, but to believe the proposition, for example, that p is more probable than not. And one believes that proposition unconditionally. At the heart of each tentative belief, there is a belief that is held with no tentativeness, although the object of that belief is not p, but a proposition that embeds p. As Newman says: "certainly, we familiarly use such phrases as a half-assent, as we also speak of half-truths; but a half-assent is not a kind of assent any more than a half-truth is a kind of truth. As the object is indivisible, so is the act" (116).


At any rate, once we get clear on the difference between subjective assessments of probability and subjective assessments of truth, the inductive argument has no tendency to show that belief is not conceptually connected with truth. The argument rests on my recognizing that there is a high probability that some of my beliefs are false. But that fact is, on my account, perfectly compatible with continuing to hold the beliefs I hold.



VIII. Doubt


The distinction between entertaining a proposition and believing it is a familiar one. One entertains the propositions representations of which are present to consciousness. We might think of the difference between entertaining a proposition and believing it occurrently as the difference between muttering a sentential representation of a proposition to oneself and muttering it to oneself while taking it to express a truth. This distinction is relevant here because it allows us to formulate in a rough way what doubt consists in. Roughly, then, one doubts those propositions one entertains but fails to believe, after making an effort to decide whether they are true or not. Of course, this leaves open the possibility that one might both doubt and believe the same proposition under different representations. But it will follow that, if one sees that they do represent the same proposition, and insofar as one is rational, one will not both doubt one and believe the other. And one cannot both doubt and believe the same proposition at the same time under the same representation; that is, one cannot both take `p' at t to express a truth and fail at t to take `p' to express a truth. Otherwise we land directly in contradiction, as we have seen. Another way to put this is that the various ways in which I can both doubt and believe the same proposition are cases of doxastic dissonance.


Note also that, in my treatment, if I disbelieve that p, that is, if believe that not-p, that is, if I take some `not-p' to express a truth, it follows that I doubt p. So doubt on my view includes both the suspension of belief and active disbelief. This is legislation, but it does capture certain ordinary-language idioms. For example, it is true that, sometimes, asked whether I doubt that God exists, I will say "I'm no agnostic. I don't doubt it, I deny it." But on other occasions, I might be asked whether, say, Fred will be back by 2:00, and if do not think he will be, I might say "I doubt it," as a way of expressing the fact that I believe he will not be.


Now one may doubt a proposition, as one may believe a proposition, occurently, tacitly, or enduringly. As with beliefs, when I refer in general to doubt simpliciter, I will usually be referring to full-fledged, enduring doubts.



The most important upshot of this chapter for the rest of this book is that belief is always verifically oriented. That is, belief is always connected with truth. One always takes what one believes to be true. This fact, as we have seen, has important implications in the treatment of such topics as self-deception. But it also has important implications for the theory of knowledge, implications I will take up in the last chapter.


1. W.V. Quine, "Reply to Kaplan" Words and Objections, eds. Donald Davidson and Jaakko Hintikka (Bostob: D. Reidel, 1975), p.342.


2. See R. Stalnaker, Inquiry (Cambridge Mass.: MIT Press, 1984).


3. Scott Soames, "Reference, Attitudes and Content," Propositions and Attitudes (Oxford: Oxford University Press, 1988), pp.197-239. Mark Richard, Propositional Attitudes (Cambridge: Cambridge University Press, 1990).


4. John Perry, "The Problem of the Essential Indexical," Propositions and Attitudes, pp.83-101.


5. Nelson Goodman, Languages of Art (Indianapolis: Hackett, 1976), pp.115 ff.


6. "I believe that . . . Mont Blanc itself is a component part of what is actually asserted in the proposition `Mont Blanc is more that 4,000 metres high." Bertrand Russell in Gottlob Frege, Philosophical and Mathematical Correspondence (Chicago: University of Chicago Press, 1980), p.169.


7. I am grateful to Jeffrey Tlumak for pointing out this problem.


8. See Daniel Dennett, "Beyond Belief," Thought and Object, ed. Andrew Woodfield (Oxford: Clarendon Press, 1982), p.18.


9. For a more elaborate account of representation in general, see my "Natural Generativity and Imitation," The British Journal of Aesthetics (January 1991).


10. e.g. in Jonathan Bennett, "Why is Belief Involuntary?" Analysis, vol. 50, #2 (March 1990), p.101.


11. "The Nature of Believing," Proceedings of the Aristotelian Society, vol.33 (1932-3), p.132.


12. Nathan Salmon, "Reflexivity," Propositions and Attitudes, p.255.


13. David Austin, What is the Meaning of `This'? (Ithaca: Cornell University Press, 1990), pp.20-23.


14. John Heil, "Believing What One Ought," Journal of Philosophy (1983), pp.752, 3.


15. This argument was brought to my attention by John Post.


16. Richard Swinburne, Faith and Reason (Oxford: Clarendon Press: 1983), chapter 1.


17. Bennett, "Why is Belief Involuntary?" p.90.


18. John Henry Newman, An Essay in Aid of a Grammar of Assent (Oxford: Clarendon Press, 1985), p.115.










Chapter II


Scepticism, Faith, and Doxastic Dissonance



Now that we are armed with rough accounts of belief and doubt, we can begin to address what I take to be the interesting questions in epistemology. That is, we can begin to ask what, if anything, we ought to believe, or, to put it another way, what, if anything, it would be a good thing to believe. The importance of the difference between these two formulations will emerge more fully in chapter four. For now, we can note that the second way of framing the matter is preferable, because it does not entail that believing is an activity; it does not entail that I can choose what to believe. In normal circumstances, where, for example, I see a truck bearing down on me, it is not up to me whether I believe that there is. So at least belief is not always voluntary. And since `ought' notoriously implies `can,' perhaps we do better to frame the question in terms not of epistemic obligations, but of good or desirable epistemic situations.


So the first question seems to be whether it is an (epistemically) good thing to believe anything at all. I take this to be the problem of universal scepticism. Now it might seem as though this is really not the question of scepticism at all as it is usually framed. In fact, scepticism about the external world, which I will take as a representative and particularly interesting example of scepticism in general, is usually framed as the question of whether we (can) have knowledge of external objects, not as whether it is a good thing epistemically to believe that there are objects external to the mind. (This alleged distinction is fundamental to Hume's treatment of the question, for example.) That is, the question seems to be, what justifies us in believing that there are objects external to the mind (or that induction is a good inference strategy, or that there are other intelligent beings etc.)?


However, as I will devote a good bit of this book to arguing, these are not different questions. Knowledge, I will argue in chapter four, is the epistemic telos; it is the purpose of believing. If knowledge were justified true belief, it would be an epistemically good thing for me to believe just those true propositions that I would be justified in believing. So the question of whether I know that there are objects external to the mind or that there is an external world is precisely the question of whether it is an epistemically good thing for me to believe that there are or is. I take it, therefore, that the question raised by scepticism about the external world can be formulated alternately as the question of whether I know that there is and whether it is an epistemically good thing to believe that there is. Thus, on my account, the question raised by scepticism about the external world is whether it is an epistemically good thing to take the proposition that there are objects external to the mind to be true.


If the issue of scepticism about knowledge were detached from the question of what one ought to believe (and it occasionally has been held to be separable), then I think knowledge would no longer be an interesting concept; no one would have any practical reason to be interested in the question of what they know and what they merely believe. If it were perfectly consistent to be an external-world sceptic and to have various beliefs about the external world, then external-weorld scepticism would be a truly useless enterprise. Now of course that is not the way Descartes, for example, conceived the relation of knowledge to belief; he entertained scepticism precisely as an intellectual hygiene, precisely to screen unjustified beliefs out of his belief system. He resolved to cease to believe what he could not show to be true; he sought to achieve a perfect coincidence of belief and knowledge. And it is only knowledge in that sense, knowledge as the goal of belief, that has any epistemic purchase in real lives.


Though most of this book focuses on relatively recent epistemological literature within the analytic tradition, I am going to move far afield from that literature in this chapter. That is because the questions surrounding scepticism are among the oldest in philosophy. And the replies to scepticism that I find most compelling, and most congenial to the overall approach of this book, are found in such figures as Diogenes, Dr. Johnson, Kierkegaard, Nietzsche, James, and Santayana. Nevertheless, I think and hope to demonstrate that the work of these figures is relevant to the current state of epistemology.



I. The Dream Argument


I am going to take external world scepticism as representative of scepticism about various matters. I think at least some of the conclusions I reach about external world scepticism can be generalized to scepticism about induction, other minds, and so forth. And I am going to take Descartes's familiar dream argument as representative of hypotheses designed to show that in fact we do (can) not know anything about the external world. Again, I think that at least most of what I shall have to say could be generalized to a discussion of the evil deceiver, the hypothesis that I am a brain in a vat, and so on.


In the first Meditation, Descartes asserts that, because he realizes that in the past he has harboured some false opinions, in order to "establish anything firm and lasting in the sciences," he has to "raze everything in my life [all his opinions], down to the very bottom, so as to begin again from the first foundations."(1) He notices that "Whatever I had admitted until now as most true I took in either from the senses or through the senses." And he points out that though his senses have occasionally deceived him about things that are "very small and distant," his sensed evidence is unequivocal that he is "sitting here before the fireplace wearing [his] dressing gown, that [he] feels this sheet of paper in [his] hands and so on." It seems that he would need some special reason to doubt that his senses inform him truly about such facts. And then he introduces the dream argument to undermine the claim that he knows what he has (seemed to) take in from or through the senses:



How often has my evening slumber persuaded me of such customary things as these: that I am here, clothed in my dressing gown, seated at the fireplace, when in fact I am lying undressed betwen the blankets! But right now I certainly am gazing upon piece of paper with eyes wide awake. This head which I am moving is not heavy with sleep. I extend this hand consciously and deliberately and I feel it. These things would not be so distinct for one who is asleep. But this all seems as if I do not recall having been deceived by similar thoughts on other occasions in my dreams. As I consider these cases more intently I see so plainly that there are no definite signs to distinguish being awake from being asleep that I am quite astonished, and this astonishment almost convinces me that I am sleeping.





The case is meant to be generalized. It is as clear a case of apparently waking experience as any could be, and it is typical. If one cannot establish in such a case that one is not dreaming, then it is hard to see in what sort of case such a thing could be established. And the argument seems to be that if one cannot establish that one is not dreaming, if one cannot justify the belief that one is not dreaming, one has no empirical knowledge whatever. In fact, this seems to be precisely the locus of the modern version of the epistemology of justification, an epistemology that it is the task of this book to refute (though the origins of the epistemology of justification extend at least to Plato's Theaetetus). That is, we are challenged by Descartes to show that we know something about the external world, or indeed to come to know something about the external world, by justifying our empirical beliefs from the ground up.


Diogenes the Cynic is supposed to have demonstrated that motion is possible by walking across a room: "[W]hen someone declared that there is no such thing as motion, he got up and walked about."(2) According to Boswell, Dr. Johnson "refuted" idealism by kicking a stone:



[W]e stood talking for some time together of Bishop Berkeley's ingenious sophistry to prove the non-existence of matter, and that every thing in the universe is merely ideal. I observed, that though we are satisfied his doctrine is not true, it is impossible to refute it. I shall never forget the alacrity with which Johnson answered, striking his foot with mighty force against a large stone, till he rebounded from it, `I refute it thus.' This was a stout exemplification of the first truths of Pere Bouffier, or the original principles of Reid and Beattie; without admitting which, we can no more argue in metaphysicks, than we can argue in mathematicks without axioms. To me it is not conceivable how Berkeley can be answered by pure reasoning. . .(3)







Moore "proved" that there are things external to the mind by holding out his hands:





I can now give a large number of different proofs [of the existence of things outside us], each of which is a perfectly rigorous proof. . . . I can prove now, for instance, that two human hands exist. How? By holding up my two hands, and saying, as I make a certain gesture with my right hand, `Here is one hand,' and adding, as I make a certain gesture with the left, `and here is another.'(4)







In this chapter, I argue that Diogenes, Johnson, and Moore take roughly the correct approach to scepticism about the external world. First, however, we need to see in detail how Descartes' initial statement of scepticism ought to be construed. For this purpose, I will make use of Barry Stroud's reconstruction of the dream argument in The Significance of Philosophical Scepticism.




II. Stroud's Version


Again, the dream argument seems to require that if one cannot establish that one is not dreaming, one has no empirical knowledge. Certainly Stroud takes it this way. One would have such knowledge, however, if one could produce a test of whether an experience was dreamed or not. Stroud writes:


If [Descartes] could find some operation or test, or if he could find some circumstance or state of affairs, that indicated to him that he was not dreaming, perhaps he could then fulfil the condition - he could know he was not dreaming. But how could a test or a circumstance or a state of affairs indicate to him that he is not dreaming if a condition of knowing anything about the world is that he knows he is not dreaming? It could not. He could never fulfil the condition. (21)


Stroud takes the dream argument's conclusion to be the proposition he terms scepticism about the external world: "We can know nothing about the external world around us." The `can' in `We can know nothing about the external world' is evidently meant to express full-fledged logical possibility. There is supposed to be a conceptual problem about empirical knowledge; it seems that the conditions on empirical knowledge are conceptually impossible to fulfil. Nevertheless, it seems to me that Stroud proceeds too quickly in the quoted passage. Surely Descartes could have an experience that would indicate to him that he wasn't dreaming. For example, let us suppose that he believed that he could only pinch himself in a waking state. Then he might take the experience of pinching himself to indicate infallibly that he was not dreaming. But Stroud argues that how ever Descartes might view the situation, the experience of pinching himself could not in fact be a good test to establish that he is not dreaming, because the "test" itself could be dreamed. "Anything one can experience in one's waking life," Stroud writes, "can also be dreamt about; it is possible to dream that one has established that a certain state of affairs obtains" (22). So what Stroud is arguing amounts to this:



(P1) In order to have any empirical knowledge, one has to know that one is not dreaming.



(P2) In order to know that one is not dreaming, one has to establish that one is not dreaming by a procedure that one knows to be reliable.



(P3) No such procedure can be known to be reliable.



(C) One can have no empirical knowledge.



Though I will frame what follows for the most part in terms of reliable procedures for establishing that one is not dreaming, I use this as a shorthand for whatever procedure Descartes or Stroud might regard as sufficient to justify such a claim. I do not mean to endorse a reliabilist theory of justification, or to impute such a view to Stroud, much less to Descartes. Nevertheless, Stroud formulates the problem in terms of reliability. In fact, that is rather odd, since it is obvious that Descartes is what the contemporary jargon would term an "internalist." That is, even were there some procedure that was in fact a reliable test of whether one was dreaming or not, Descartes could not count the performance of such a test as establishing that one was not dreaming, unless one could establish, on the basis of one's incorrigible internal states, that the test was indeed reliable and had indeed been performed. I will have much more to say about this sort of internalism in the next chapter, but right now I will proceed on the (false) assumptions that Descartes' internalism is true, and that Stroud's talk of "reliability" could be rendered compatible with Descartes' overall picture of justification.



III. Qualifying Scepticism


Now it might seem that (P2) is too strong, that all Stroud needs is the weaker claim that in order to know we are not dreaming we need to establish that we are not dreaming by some reliable procedure. But notice that Stroud does not in fact deny that there could be such a procedure. And surely, it is in principle possible that there could be.(5) For example, perhaps in fact I never dream that I am pinching myself. Then pinching myself would be a perfectly reliable procedure for establishing that I am not dreaming. (Again, this would be radically insufficient for Descartes to count the belief as justified. But it seems to be all that is required on Stroud's recapitulation of the dream argument.) Nevertheless, I could not establish the procedure itself to be reliable, because I would hardly be in a position, without begging the question, to establish that I never or rarely dream that I am pinching myself. That is, I could not establish the reliability of the procedure without assuming that I am not dreaming as I attempt to establish that I never or rarely dream that I am pinching myself. And that is Stroud's point.


So again, what is being claimed is that one only has knowledge where one knows the procedure for arriving at a belief to be reliable. The reliable procedure justifies the belief, on Stroud's view. This seems, however, to involve all knowledge-claims directly in a regress. The procedure must be known to be reliable. This in turn means that it must be established to be reliable by a reliable procedure. Furthermore, again, that procedure must be known to be reliable. And so forth. If it is a requirement on knowledge that it be established by a procedure that is known to be reliable, it should hardly be surprising that we do not know anything at all.


Stroud is evidently relying on the familiar account of knowledge as at least justified true belief, and he is evidently appealing to procedures and tests as strategies for justification. But if knowledge is at least justified true belief, Stroud's discussion involves such a strong view of justification that it immediately makes knowledge impossible. Stroud does not, in other words, even need the dream argument in order to establish the sceptical conclusion. If knowledge is true belief established by a procedure that is known to be reliable, any knowledge-claim involves the claimant in a regress. We should surely not take this to show that knowledge is impossible, but merely that we should not give a conception of justification that immediately entails a regress, if we can give a plausible conception that does not.


Thus, let us consider Stroud's argument with altered second and third premises:



(P2.1) In order to know that one is not dreaming, one has to establish that one is not dreaming by some good argument or reliable procedure.



(P3.1) No such procedure can in fact be produced.



Here we do not require that the procedure or argument itself be known to be reliable. Note, however, that in such a case it is perfectly possible to know that one is not dreaming if some reliable procedure can be produced. And as Descartes and Berkeley among many others have pointed out, we do have extremely familiar ways of distinguishing dreams from waking experience, such as noting the continuity of the latter in comparison with the former.(6) (However, we have no familiar procedures for establishing that we are not being deceived by a malin genie or that we are not brains in vats. And these tests would themselves be question-begging if we took a strictly internalist stance about justification.) On the revised argument, the question seems to turn on whether such procedures are in fact reliable, not on whether they can be established to be reliable. Stroud's argument, in other words, relies on what William Alston calls a "level confusion."(7) The revised argument allows that it is perfectly possible to know that one is not dreaming; that is, it is perfectly possible that there is a reliable procedure or test. Let us admit, with Stroud, that is is always possible to dream that one has applied such a test. Then the situation would be as follows: it is perfectly possible to know something about the external world, but it is not possible to know that we know it, or to establish that we know it. Thus, by the lights of the revised argument, knowledge about the external world is hardly conceptually impossible. But the situation just described is itself problematic, for, on it, we cannot know when we know and when we merely dream that we know. And we should note once again, that no level confusion occurs if we take a strictly internalist view about justification. Then everything does turn on whether the procedure can be established to be reliable by evidence internal to the doxastic agent. So on an internalist construal, the dream argument stands in its traditional form.



IV. The Beginning of a Reply


Let us treat external world scepticism, then, consistently with our revised version of Stroud's argument, not as the position that empirical knowledge is logically impossible, but as the position that no particular claim to empirical knowledge can be established. (I think the sort of argument I will recommend will apply mutatis mutandis to the traditional problem as formulated within internalism.) Then the question I want to press is this: do we need to establish that, in general, some such claims are true? In other words, second-level doubt about knowledge of the external world easily enough infects first-level belief about the external world. If we cannot establish when we do know something about the external world, it might seem that the only rational approach is to suspend judgement on all claims about it.(8) Such claims can, perhaps, be established, if we take an externalist approach to justification. But if Stroud is right, it is impossible to know that they have been. So the sceptical dilemma in something like the traditional sense persists.


I now want to turn to a reply to scepticism about the external world in this sense. It is not a reply in the style of Descartes or Kant or Carnap, but in the style of Diogenes, Johnson, and Moore.


I cannot know a proposition unless I believe it.(9) Belief, as we have seen, can be treated as a propositional attitude. Likewise, to doubt some proposition implies at least this: that one fails, while doubting it under a certain representation to believe it under the same representation. (That is, if x is minimally rational, ~(Bxpr & Bx~pr). Of course, it is necessarily the case that ~(Bxpr & ~Bxpr). The other possible case of doxastic dissonance, Bxpr & ~Bxps, is not in play here.) Sceptics raise doubts about the external world, and the project of refuting scepticism amounts in part to giving reasons why one should not doubt all claims about the external world. Now again, I have admitted that I cannot show that I know I am not dreaming right now. But as a matter of fact I am not in the least worried that I may be dreaming right now. I do not harbor the least doubt that there is an external world. So I do not see what role a refutation of scepticism in this sense could play in my mental economy. It could not possibly make me more confident than I already am that I have empirical knowledge, so I do not see why, if it is possible that I do, I should try to establish that I do. When Diogenes walks across the room and Johnson kicks the stone, I take it that they have some such point in mind. They are saying, I think, that we do not in fact suspend judgement about the reality of motion or of the external world.


Now that might well seem to be a feeble reply to philosophical scepticism, or worse, it might seem not to be a reply to properly philosophical scepticism at all. Stroud (and many others) would simply argue that I, and Diogenes et al. insofar as they do what I represent them as doing, have confused two very different questions: an "internal," practical question and an "external" philosophical one. He writes:


What is seen to be true from a detached `external' standpoint might not correspond to what we take to be the truth about our position when we consider it `internally,' from within the practical context which give our words their social point. Philosophical scepticism says that the two do not correspond; we never know anything about the world around us, although we say or imply that we do hundreds of times a day. (81)







Notice that the difference between the internal and external standpoints is not what is true from each standpoint, but what is taken to be true. Stroud does not want to hold that one and the same proposition is true from one standpoint and false from another. And notice also that Stroud admits that our verbal practice does show that we take ourselves and imply ourselves hundreds of times a day to be committed to the claim that we know something about the external world.


Stroud gives the following analogy to elucidate the internal/external distinction. During wartime, airplane spotters are taught to identify a plane as an F if it has features x, y, and z. Unbeknownst to the spotters, Gs also have those features, but the distinction between Fs and Gs is unimportant for the war effort. So though they believe they know that a certain plane with x, y, and z is an F, they know no such thing, as becomes evident from a point of view wherein one is in full possession of the facts. This point of view is what Stroud refers to as "external" to the practice of airplane spotters.


Now I want to allow that this analogy is legitimate, as far as it goes. It is possible that from a position in which I am in full possession of the facts, it would be evident that what I take to be the external world consists in my own representations. (However, I am going to raise various problems about "internal representations" in the next chapter.) I take this to be possible because I agree that we cannot establish that we are not dreaming without begging the question. And it is certainly possible to believe something emphatically and not know it; if the claim is false, no one knows it to be true.


The spotter believes that the plane is an F. Let us stipulate that the plane is not an F. Does what he believes contradict the facts? Of course. It is not the case, as Stroud sometimes hints, that the belief of the spotter is true internally, or true within the current practices of airplane spotters. The belief of the spotter is false simpliciter, false by the light of the facts. And as Stroud asserts (72), the spotter could be made to see that it is false. Seeing that a claim is false (under a representation), insofar as one is rational, involves among other things ceasing to believe it (under that representation), if the theory of belief developed in the last chapter is right. So in this case we have a perfectly clear opposition between the "internal" and the "external" points of view.


And Stroud does admit that an assertion made from the internal point of view can contradict one made from the external point of view (126). For the issue turns in part on whether there is in fact an opposition between the internal practical claim, for example, that I am walking across a room or kicking a stone and the external, philosophical claim that I might be dreaming. (I will take the claims that I am walking across the room or kicking a stone to have much the same status that Stroud attributes to Descartes' experience by the fire; they are typical cases of assertions that entail that there are external objects, and the particular properties of these experiences are meant in a certain sense to be ignored.) The internal claim that I know that I am walking across a room entails the internal claim that I know I am not merely dreaming that I am doing so (or else the dream argument itself does not go through). Then that internal claim seems to contradict the external claim that I do not know I am not dreaming. If all that is meant by "internal" and "external" is that from the external point of view we take into account information that is unavailable to the internal view, there is no reason at all to think that the claims themselves that I know I am not dreaming that I am walking and that I do not know it do not contradict one another. In fact, if I do believe both of these things, then a both believe and disbelieve the very same proposition under the very same representation: Bxpr & Bx~pr. This is obviously an undesirable doxastic dissonance, one that, as I have argued, I will try to adjudicate if I become aware of it. Scepticism would then be interpreted as an attack on our everyday claims to know, and of course that is just how Stroud and Descartes do interpret it.


Stroud asserts, however, that, in the case described, the spotter might be justified in claiming to know that the plane is an F. That just means that in the situation he finds himself in, the claim that the plane is an F is reasonable. But of course it can be reasonable to claim that one knows something in certain circumstances even where one does not know it. That has no tendency to show that claims to know in the internal sense do not contradict external doubts about knowledge.


If the spotter is rational, then when he comes to know that Gs have x, y, and z, he will come to doubt that the plane is an F. (It is likely, however, that the spotter is judging probabilities, and that, for example, if there were extremely few Gs, he might remain confident that the plane was an F. Let us suppose that the spotter believes that the probability that a plane he spots with x, y, and z is an F is roughly equal to the probability that it is a G.(10)) That is to say, among other things, that he will cease to claim to know that it is an F, in the absence of compelling motivations to make a claim he knows to be false. In the absence of such motivations, the new information rationally requires him to recant his claim to know that the plane is an F. But as long as he sticks to it that he knows it is an F, how ever he comes to believe that, he cannot consistently doubt that it is an F, or think it could well be a G. That is, to do these things would be to allow himself to retain his doxastic dissonance. Of course, it is perfectly possible to have contradictory beliefs. It is not a contradiction to say of someone that he believes that he knows P is an F and that he believes that he does not know that P is an F. One may believe and disbelieve the same proposition under different representations. For example, when the proposition is framed in sophisticated philosophical jargon one may take it to be true, and when it is framed in everyday language one may take it to be false. And one may also get into the position of believing and disbelieving the same proposition under the same representation at different times. For example, one might believe that one does not know that there is an external world while one is lecturing or writing a book, and believe that one does know it while one is cooking dinner or changing the baby's diaper.


One is in a state of doxastic dissonance where one believes a proposition under one representation and disbelieves it under another. And one might also be said, in a slightly extended sense, to be in a state of doxastic dissonance where one alternates between believing and disbelieving a proposition under a single representation. (Of course, one's beliefs change over time, and the fact that one believes p at t and ~p at t1 does not entail that one is in a state of doxastic dissonance. But if one alternates back and forth between p and ~p under the same representation over a relatively short duration, one is experiencing doxastic dissonance.) I take it that it would be desirable, if possible, to render our beliefs consistent. I take it, that is, that doxastic dissonance is in general something we would prefer to avoid. And I take it to be a fact about human beings that they seek to reduce doxastic dissonance, that they try to render their belief-sets consistent. And in fact reduction of doxastic dissonance is one of the goals of philosophical and other sorts of inquiry. (More precisely, it is one of the most obvious ways of pursuing the overarching purpose of inquiry, which is, as I will argue in the last chapter of this book, to believe the truth.)


Again, as we have seen, the "external" claim that I do not know I am not dreaming contradicts the "internal" claim that I do know that. Thus, the person who believes the one and disbelieves the other (perhaps because the external claim is represented in the jargon) is in a state of doxastic dissonance, as is the person who believes the external claim while writing or lecturing and disbelieves it while cooking dinner or changing the baby. Likewise, the person who believes "internally" relative to certain considerations that he knows he is not dreaming, and "externally" relative to others that he knows he might be, or that he cannot know he is not, is in just such a state. This person believes for everyday practical purposes that he is not dreaming, and, as Stroud points out, affirms that belief hundreds of times a day by producing utterances like "Here's dinner." But for "philosophical," "detached," "objective" purposes he believes that he might be dreaming. The propositions he believes directly contradict one another, though different considerations might make each seem compelling at different times. The person who believes all this is, I think, undergoing internal conflict.


Now I would like to ask whether it is easy or difficult to entertain "external" doubts about the existence of the external world. To read Descartes, or Stroud, one might get the impression that nothing could be easier. Stroud even seems to believe that it is natural or inevitable to entertain such doubts. But notice that, if Stroud is right, such doubts contradict the kind of claims we make hundreds of times a day. The fact that we do affirm hundreds of times a day things that entail that we are not dreaming, shows that we are already committed, massively committed, in our practice, to the falsity of philosophical scepticism. We commit ourselves to this view over and over again without the slightest hesitation. Refraining from committing ourselves to it would require constant exertion over the course of an entire lifetime. We would be in the position of attempting to withold belief from just those claims we seem to ourselves most clearly to know. If one achieved doubt of this sort, one would presumably seek a refutation of scepticism. But doubt that does not arise cannot be assuaged.


Sceptics and those who take scepticism seriously often enough appeal to a distinction between `philosophical' and `practical' beliefs, between `methodological' and `genuine' doubts, between `internal' and `external' viewpoints. I am arguing that any such distinction stands desperately in need of clarification. How can I hold a belief philosophically that I do not hold at all, and what methodological advantage is supposed to accrue from making making assumptions that I am absolutely sure are false? This is not like acting as though some hypothesis is true, in order to test it; there, the hypothesis must be a candidate for belief, and the more plausible the better. If we make the distinction the way Stroud does, and simply specify that more information is available from the external than the internal point of view, then, if there is a conflict and if I am rational, I will simply jettison the internal point of view. If, on the other hand, we say that philosophical doubts are doubts we profess but do not entertain, philosophical beliefs beliefs we mouth but do not hold, then I simply do not see in what sense these are doubts or beliefs at all, or why we should bother trying to assuage or exorcise them. So I am arguing that if we are already sure that we have empirical knowledge, there is no point is trying to establish that we do in a very general way. If scepticism is an attack on our everyday claims to know, then it is obviously not merely `philosophical,' `methodological,' `external.' And if it is not an attack on our everyday claims to know, what is it an attack on?


Some philosophers profess to have achieved doubt about the external world. But that no more shows that they have achieved doubt than the claim of the spotter to know that the plane is an F shows that he knows it is an F. If such philosophers really have achieved doubt about the external world, it might be massively dissonant with their other professed beliefs, and without the greatest mental discipline such doubt could not then be sustained. The demands of their own lives, I mean their "internal" practical lives, their lives as particular creatures, would need to be constantly suppressed. And if, as a philosopher, I get into the position of professing to doubt what I am sure of all along, if with the utmost seriousness I raise doubts on paper that I do not in fact entertain, I am open to the charge not simply of making false claims, but of being a hypocrite. My beliefs are, to repeat, expressions of my emotional life. The process of universal doubt is an attempt to purge belief of its emotional component, to render belief "purely rational." First of all, we might get worried here about the purported opposition between emotion and rationality. But second, if I am right about belief, the attempt to purge belief of emotion is an attempt to purge the agent of belief entirely. That is not a "methodological" exercise.







V. Moore's Proof


In his "proof of the external world," Moore takes himself to be trying to demonstrate conclusively that there are things external to the mind. That is, Moore puts forward his argument as a direct refutation of scepticism. If Moore knows that here is a hand and here is another, then it follows that Moore, if he reflects on the entailments of that belief, knows that he is not dreaming that he sees two hands. Thus he will have established that he is not dreaming. But whether Moore knows the former - that here are two hands - is precisely what is at issue. The sceptic holds that because he could in fact be dreaming that he is seeing two hands, he cannot establish that he knows that here are two hands. So in the absence of further argument it does not follow that Moore knows he is not dreaming. This is the gist of Stroud's answer to Moore, and as an answer to a direct refutation of scepticism it strikes me as beyond reproach.


Moore's challenge, however, might be read in another way. By holding up two hands, we might say, Moore is dramatizing to his audience (the paper was delivered as a lecture), that they in fact do believe that here are two hands.(11) Moore asserts that this sort of belief is more firmly held by these particular persons (and we may take ourselves to be part of Moore's extended audience) than any belief that could be produced to attack it.(12) If you believe with this kind of confidence that you know that here is one hand and here is another, and acknowledge that the proposition that you know here is one hand and here is another entails the proposition that you are not dreaming (which Stroud certainly admits), then you ought not to be concerned any longer with scepticism about the external world. If I am dreaming that here are two hands, I do not thereby know that here are two hands. But if I believe that here are two hands, and I follow out the implications of that claim, I am rationally committed to affirm that I am not dreaming. This is also the way I want to read Diogenes and Johnson: as showing us what we do believe, where our genuine commitments already lie.


Marie McGinn, in her book Sense and Certainty, also reads Moore this way. She writes: "Moore is using the fact that scepticism is simply beyond belief, and the fact that it is in conflict with what we are all absolutely convinced of, in an attempt to rout the sceptic."(13) And McGinn, as I do, attacks Stroud's distinction between internal and external viewpoints as an unsuccessful attempt to insulate philosophical from actual doubt. Nevertheless, McGinn attacks Moore's proof as follows:



[I]t seems to me that Moore's dogmatism is, by the principles of the practice he means to defend, an unacceptable response to a demonstration that one lacks the requisite justification to sustain a knowledge claim. It is impossible to use our conviction that the sceptic's conclusion is false as a legitimate ground for dismissing scepticism, for it is clearly part of our ordinary grasp of the concept of knowledge that personal conviction is never sufficient to warrant the affirmation of a knowledge claim exposed as doubtful. Possessing an adequate justification for believing a given proposition, p, is part of our normal ground for asserting a claim to know p. (p.50)



As the reader is aware, I will argue that possessing an adequate justification for believing a proposition is not part of our normal ground for asserting a claim to know (though I also deny that personal conviction warrants knowledge claims; external-world truth is required). And either way, this argument is surely without bite at this point, because part of what Moore is doing is precisely trying to show that, in the most run-of-the-mill cases, justification is irrelevant to knowledge claims.


McGinn continues:



Thus, aside from going against our ordinary understanding of the words `I know,' Moore's proof is unsatisfactory in so far as it does not earn our commitment to common sense, by means of a philosophical understanding of why the impossibility of grounding our framework judgements does not constitute a threat to, or demonstrate a failure of, our ordinary practice. A genuine resolution of the riddle of scepticism must allow us to see why our inabiility to justify the judgements that represent our normal starting point is not a lack of anything essential. (p.53)



I hope to satisfy McGinn's demand in what follows in this sense: I hope to show that our inability to ground some particular judgement, or some particular sort of judgement, would not demonstrate a failure in our ordinary practice. I think, in other words, that McGinn is far too credulous in identifying ordinary practice with a particular, optional philosophical orientation: the orientation I have called the epistemology of justification. And I also want to remark that McGinn's own solution is unsatisfactory. Following Wittgenstein, she treats Moore-type propositions as part of the "framework," as logical rather than empirical, the ground rather than the result of empirical investigation. I will return to these points briefly in the next chapter, but for now I point out, first, that the claim that "here is a hand" is empirical; it is about the world, not about language. And second, this view represents an invidious verificationism; it will automatically exempt from the realm of the empirical any claim which cannot be empirically justified. This verificationist orientation is both unsatisfactory in its own right and question-begging in this context.


If one does not think that one can know such claims as that here are two hands to be true, because one might be dreaming that here are two hands, I, like Stroud and McGinn, do not think that one faces a refutation. What I think one faces, rather, is the prodigious labour of renouncing the world by suspending judgement as to the existence of each item one seems to find there. That is, one faces the prospect of resolving one's doxastic dissonance by giving up many of one's most firmly held beliefs. This is a personal rather than an epistemological crisis. Let us admit that, other things being equal, it is rational to believe what is entailed by one's other beliefs. And let us admit that scepticism about the external world rationally obliges the sceptic to suspend judgement about such claims as that here is a hand and here is another. Then the task of suspending judgement about each such claim is incumbent on the rational sceptic.


Again, Stroud sets out the difference between the internal and the external points of view in terms of the information available in each perspective: the external view includes all the information available in the internal view and more besides. (Recall the spotter case; there from the external view we still know that the plane has x, y, and z, but we know also that Gs have x, y, and z.) The internal view fails to take into account the supposed fact that one might be dreaming, while in the external view that supposed fact is taken account of while the experience retains its phenomenological character. But, as Stroud maintains, it does not follow directly from the fact that he claims that he does not know that he is not dreaming in the external point of view that the philosopher has introduced a new or overly rigorous notion of knowledge. Stroud emphasizes that the philosopher is using just the everyday notion, and pointing out that though it is appropriate to assert from the internal point of view that one knows that here are two hands, it is still a question from the external point of view whether this case meets the requirements on knowledge or not. That is just to admit, however, that one who is looking at the matter internally or practically is rationally obliged to disagree with the claim that he does not know he is not dreaming. However appropriate it is to assert that one knows that here are two hands, the sceptic holds precisely that one is obliged to show that one knows this.


That is why scepticism is not merely an epistemology, but a way of life, as the Pyrrhonians insisted. One who is indeed able to cultivate that way of life has my esteem for his constant philosophical diligence in the face of his own everyday experience. But I admit that the predicament of such a philosopher strikes me as slightly comical as well. It seems to me that the world discloses itself to me over the decades of my life. World and organism achieve an exquisite mutual adaptation. My senses and intelligence are alert to the world's vicissitudes, and the world rewards my alertness more often than not by providing me with some coherent experience. (I will fill in this picture of human experience in the next chapter.) I am afraid that one who cultivates a life of scepticism would maintain a detachment in the face of such experience that I would find intolerable. Of course, the Pyrrhonians thought that the ability to achieve such detachment was precisely the desirable aspect of scepticism as a way of life. I am not going to get into a debate about how we ought to live; I merely repeat that I would find such detachment intolerable. But what gives the position of the sceptic its faintly comic flavor is the fact that the most concrete facts of life are renounced in favor of the flimsiest abstractions.


I think, however, that such cases must be rare, since experience of the sort I have when I see two hands, insofar as it constrains me to affirm that here are two hands, itself stands as a constant objection to scepticism. Of course, ascetics and mystics of various sorts do renounce the world in various ways. I am by no means denying that it is possible for persons to suspend belief about the existence of things external to the mind.(14) I think the position of such persons is unassailable as far as decisive argumentation or reliable procedures or any other sort of justification for the denial of their views is concerned. But I do not admire a philosopher who claims for the purposes and length of a book that he does not know whether there are things external to the mind and then turns away from the page to say "here's dinner." Such a philosopher does not believe what he is writing, or else he does not believe what he is saying. And the truth of the matter is that he does not believe what he is writing.


Stroud begins his book by saying that "scepticism as a way of life is not my subject" (viii). Rather, he is concerned with scepticism as a philosophy. I am arguing that Stroud has not shown us how to distinguish philosophies from ways of life, and I am asserting that there is no such distinction.


VI. Faith


I will certainly be accused of having left epistemology behind completely and entered into psychology. For the discussion seems to have turned from what we ought to believe or what it would be a good thing to believe to what we in fact believe. It will be said that it is obviously true that the genuine commitments of most people, and of most philosophers, show that they believe that there is an external world. But it will be asserted that the characteristic question of epistemology is whether such commitments are or are not justified. This is taken to bear on the question of whether we know that there is an external world, because justification is taken to be a component of knowledge. Now I will argue in the last chapter that justification is not a component of knowledge, that knowledge is merely true belief. But before offering that argument, I would like to explore in somewhat more detail the notion of unjustified belief, or, as I will call it, faith. For if it turns out that we cannot do without faith, this in itself would provide an argument that we should not try to.


The argument of the next several sections is a generalized attack on the epistemology of justification. As I have pointed out, the epistemology of justification is historically associated with "methodological doubt." This is because the assumption that methodological doubt is legitimate leads to a picture in which we need, in order to have any knowledge at all, to start justifying our beliefs from the ground up. It is this view that extends directly to the contention that justification is logically required for knowledge (though there are other ways in which this contention has arisen). But as I argued in the first chapter and in the preceeding sections of this chapter, the notion of a purely methodological doubt is not a notion of doubt at all. That is, to doubt some proposition is either to disbelieve it under some representation, or to entertain it and fail to believe it under some representation. Thus, all doubts are somebody's doubts; all doubts are attitudes somebody really has toward some proposition. What follows constitutes an attack on the epistemology of justification by pressing the question of who is in fact capable of doubting what, and by pressing an attack on methodological scepticism.


To repeat, I use the term faith as synonymous with unjustified belief. This will annoy some specialists on the subject, and certainly some who use faith as a fine-grained theological notion. I think that even in the theological discussion, the notion of faith often includes the unjustifiedness of the articles of faith (though it does not always include that, e.g. in Aquinas). But I do not offer the notion of unjustified belief as an exegetical tool for any particular discussion of faith; if the reader prefers, she can regard my use of the term as a sheer stipulation. My argument, however, will be that faith in the present sense on certain matters is extremely difficult and not desirable to avoid. It will follow that, if what is ultimately desirable doxastically is knowledge (as I will argue later), faith can rise to the level of knowledge. If faith can rise to the level of knowledge, then justification is not logically necessary for knowledge.


My discussion here leaves the contemporary debate more or less completely, though I think it bears directly on that debate. The notion of faith as I want to make use of it here in general epistemology was developed most elaborately by Kierkegaard, James, and Santayana. So what follows is an attempt to bring these figures more fully into the contemporary debate about knowledge.


Kierkegaard characterized faith as an objective uncertainty held fast in passionate inwardness.(15) I suggest that this is roughly what James meant when he he described faith as a believing attitude that is taken up in spite of the fact that our "merely logical intellects" may not have been coerced into the attitude in question.(16) That is, to have faith is roughly to believe in the absence of justification. Note that both Kierkegaard and James emphasize the emotional aspect of faith, an aspect which, if the account in chapter one is correct, is present in all belief to one degree or another.


To begin with, I would like to explore the use of the notion of faith in Kierkegaard and Santayana in their criticism of what I have been calling the epistemology of justification. Such an epistemology holds that if one is to know p one must be justified in believing p. It is possible to read Plato, in the Theatetus and elsewhere, as expressing such a view (though that is not how Kierkegaard would read Plato, and it should be noted that Socrates does appear to reject this view by the end of the dialogue), and certainly it appears full-blown in Descartes. In less technical terms, the epistemology of justification holds that we ought to believe just what it is rational for us to believe, and that all assumptions must be subjected to rational scrutiny. Such a view informs the writings both of the rationalists and of the empiricists. Rationalists and empiricists converged on a search for indubitable and objective knowledge, and the former thus took a detour through scepticism (Descartes), while the latter often remained there (Hume).


The epistemology of justification, epistemology conceived as an abrogation of scepticism, was rejected by both Kierkegaard and Santayana for roughly the same reasons. And these reasons turn on the notion of faith. To Kierkegaard, who had a supernatural view of man and the cosmos, the epistemology of justification seemed a betrayal of spirituality, of the ultimacy of subjective experience, and of the emotional core of human belief. To Santayana, a naturalist and materialist, it appeared to be a misguided and finally impossible attempt to escape our own animality. But both men found at its heart the hypocrisy I discussed in the last chapter. Doxastic dissonance, when it becomes militant, reaches the level of hypocrisy; it insists that its proponent believes what he in fact fails to believe, or fails to believe what he in fact believes. It tries to lop the rational from the emotive aspects of human experience. Both Santayana and Kierkegaard found in the epistemology of justification an attempt to leave behind what no person can leave behind and remain a person. If Moore tried to show us where our real commitments already lie, Kierkegarrd and Santayana try to give us an account of why, as persons, our real commitments do lie there. And both Kierkegaard and Santayana tried to show why we need not and should not violate those commitments in philosophy, just as we need not and perhaps even cannot violate them in our everyday lives. Another way of putting the point is again something we concluded above, namely that there is no distinction between philosophies and real ways of life. The epistemology of justification divides us from ourselves.





VII. Faith and Scepticism


In one of his early works, Kierkegaard terms faith the opposite of doubt.(17) By this he does not mean simply that faith overcomes doubt (though of course in certain circumstances it does), nor that faith is a form of certainty while doubt is a form of uncertainty (though of course that is true, if certainty is construed as a subjective state). Rather, in Johannes Climacus or, De Omnibus Dubitandum Est, Kierkegaard is concerned with the question of whether philosophy begins with doubt; he is concerned with the question of whether the philosopher can begin with doubt, and, if so, whether and how he proceeds after he has doubted. That is, the attack encompasses methodological doubt. Likewise, Kierkegaard was concerned with the question of whether it is desirable to arrive at faith, and, if so, by what route one arrives there. In other words, the opposition between faith and doubt was for Kierkegaard a philosophical opposition. But nothing that for Kierkegaard was worth writing about could be said to remain a merely philosophical question, to be pursued only academically. Rather, the opposition of faith and doubt is philosophically important because it confronts us with a momentous choice: where and how to begin the process of understanding our own intellectual lives.


We have seen that Descartes' sceptical doubts about (almost) all purported knowledge form a starting-point of a certain conception of epistemology, a conception which continues to this day not only in the writings of those who, like Stroud, profess a sympathy with scepticism, but also in the writings of all those who assert that every case of knowledge is a case of justified belief. Implicit in all such views is the notion that we know only where we are in a position to eliminate or mitigate doubt. And here this does not mean the achievement of psychological certainty, but rather being able to give reasons for belief, considerations that would effect the reasonableness of anyone's accepting the claim in question. In fact, it is just this picture that gives the method of universal doubt its bite, though the picture survives the method. If the philosopher begins with doubt, it is supposed to become possible to see in which cases good reasons for belief can be given, and that is taken to bear on the question of whether and where we can have knowledge.


Descartes had no sooner formulated his doubts than he began to get beyond them, and it was in getting beyond doubt that modern epistemology found its particular project. German idealism as expressed by Hegel represents for Kierkegarard the culmination of this assuagement, of the philosophical easing of philosophical doubts. According to this view, again, the discipline of epistemology consisted in an examination of the conceptual (as opposed to the psychological) bases of our beliefs: the attempt to show how reason could extricate us from doubt, or (as again with Hume) to emphasize our inextricability. We could, in other words, come to see what knowledge (if any) we possess, by this technique: doubt everything, or at least everything that we can find a reason to doubt or a strategy for doubting, and see to what extent justificatory activities can remove our doubts. Where such activities indeed remove doubt, there we have knowledge.


And again I want to suggest that, though one no longer finds philosophers engaged in the task of doubting everything that can be doubted, epistemology is still dominated by this picture of its project. When philosophers argue over whether a belief is justified in virtue of its coherence within an explanatory structure or rather in virtue of its relation to indubitable or prima facie justified basic beliefs, they are not only arguing about justification. They are arguing about what it would be a good thing to believe. What connects them to the tradition that Kierkegaard and Santayana reject is the assumption (it is very rarely more that that) that these are precisely the same thing, that it is a good thing to believe p if and only if one is justified in believing p, or in a position objectively to remove doubt about it. Thus the task of universal doubt as conceived by Descartes remains relevant to contemporary epistemology.


This task, however, entered into with such blitheness and optimism, may be far more difficult than it appears. For it is precisely we who purport to have knowledge, and, even more fundamentally it is precisely we who seem to start on the road to knowledge by doubting everything, or, on the contemporary project, by subjecting all our beliefs to rational scrutiny for the purpose of justifying them, for the purpose of believing only what we are justified in believing. That is, this epistemological inquiry is necessarily an inquiry about human subjects, about wherein human subjects are capable of achieving knowledge. Kierkegaard and Santayana thus begin by pushing the epistemological inquiry back one step. They ask, not "Are we capable of knowing anything?" but "Are we capable of doubting everything?". Descartes' casual doubts, entertained as he sat by the fire and toyed with sceptical hypotheses such as that he might be dreaming and that a malin genie might be bent on deceiving him, were they genuine doubts? That is, did he succeed in actually doubting what he described himself as doubting? As I have argued, there are good reasons to think that Descartes did not succeed in doubting what he professed to doubt. To repeat, to doubt the existence of the external world is a way of life and not only a philosophy. Furthermore, it is a way of life that demands an extremely rigorous and perverse approach to each of one's experiences of the external world. Doubting the existence of the external world is, as Kierkegaard points out, is no frivolous activity for an idle moment.


In Fear and Trembling he expresses himself on the matter with typical acerbity:



Every speculative price-fixer who conscientiously directs attention to the significant march of modern philosophy, every Privatdocent, tutor, and student, every crofter and cottar in philosphy, is not content with doubting everything but goes further. Perhaps it would be untimely and ill-timed to ask them where they are going, but surely it is courteous and unobtrusive to regard it as certain that they have doubted everything, since otherwise it would be a queer thing for them to be going further. This preliminary movement they have therefore all of them made, and presumably with such ease that they do not find it necessary to let drop a word about the how; for not even he who anxiously and with deep concern sought a little enlightenment was able to find any such thing, any guiding sign, any little dietetic prescription, as to how one was to comport oneself in supporting this prodigious task.(18)





As he does in many places in his writings where this issue surfaces, Kierkegaard goes on to describe universal doubt as something that could be achieved only by a lifetime's exertion, and perhaps not even then. And were one to make this exertion, to spend a life in pursuit of doubt, one would have no time or energy or ability left for the equally prodigious task of extricating oneself. So here we have a more or less systematic motivation for the claim made above that scepticism must be regarded as a way of life. And the implications of that claim for contemporary epistemology of justification are just as clear; it is an extremely rigorous task to believe only what we have good reasons to believe, and hence we had better make sure before we embark on this task that we find it desirable.


Now as we have seen, the practitioner of "methodological" doubt might reply that it is not as an individual human being that he doubts everything (otherwise he could hardly also continue to live and philosophize), but that as a philosopher he finds it necessary to entertain such doubts in order to clarify just what it is he does know. This would be to apply Stroud's "internal" and "external" points of view to human beings, or their capacities. That is, when I take up the "external" point of view I am acting in my capacity as a philosopher, while in the "internal" point of view I am acting in my capacity as an agent, an active human being. Hume, raising doubts about the supposed necessary connection of cause and effect, wrote as follows: "My practice, you say, refutes my doubts. But you mistake the purport of my question. As an agent I am quite satisfied in the point; but as a philosopher, who has some share of curiosity, I will not say scepticism, I want to learn the foundation of this inference."(19) The advocate of the epistemology of justification may well say something like the following. "Of course, as particular agents, it is unlikely that we are going to be able to believe only what we are justified in believing, but as philosophers, we postulate this as an ideal, and examine the actual state of agents with that ideal in mind." But we need to ask: why should this theoretical, philosophical ideal have anything to do with how we actually conduct our lives? If the doubts which are raised are genuine doubts, then it seems that, as in the case of Hume, they introduce doxastic dissonance into the epistemic agents who have those doubts. If, on the other hand, they are not genuine doubts, they ought not to be taken seriously.


Now this distinction between agents and philosophers, or, alternately, between internal and external viewpoints, is just what Kierkegaard is puzzled about. When one acts qua philosopher has one ceased to act qua agent? Has one, in Kierkegaard's terms, ceased to be a human being and become, say, philosophy itself? That would be an incredible metamorphosis. And as we have seen, even supposing that such a metamorphosis were possible, it would divide a person against herself; it would introduce systematically a deep doxastic dissonance with regard to a whole series of propositions, since it is the very same propositions which we approach as agents and philosophers, from the internal and the external standpoints. And it would divide one's cognitive from one's affective faculties. As a philosopher, Hume doubts the necessary relation of cause and effect; he doubts it from the "external" point of view. But as an agent he all the time believes it. As we have seen reason to believe, the claims in question simply contradict one another. So Hume is in the business of cultivating quite consciously his own dissonant state, of denying what he believes and believing what he denies. That is, and ought to be seen as, perverse. Or, and perhaps more plausibly, Hume simply professes doubts which he does not hold. If the doubts raised by Hume about the necessary conection of cause and effect are indeed compelling, then he ought to get busy doubting the necessary connection of cause and effect not only in his books, but in his life.


But notice that, for Descartes (though not for Hume), the idea is not to raise doubts that cannot be assuaged. The project of methodological doubt is designed to set knowlege on a firm foundation; we entertain doubts because we need to do so in order to become certain. But it is Kierkegaard's point that, if one enters deeply and earnestly into doubt, one will find it impossible to begin to philosophize one's way out of doubt. If one were to genuinely doubt the existence of the external world, according to Kierkegaard, one could not in fact establish that it exists. And I have already agreed with that; I do not think that the dream argument can be refuted, only I am claiming that it need not be. It is certainly possible to view Descartes' own attempts to extricate himself from doubt as either abject failures or as more or less wholly disingenuous. That is, his "deduction" of God's existence from his own ideas is notoriously fallacious. And it is a rather odd coincidence that Descartes ends up more or less exactly where he began: believing he is sitting by the fire, believing that the Christian God exists, and so forth. Here it is possible to see a dance of rationalization rather than of rationality. Descartes winds up generating arguments, more or less unconvincing, for positions to which he never ceased to be passionately committed. But it is in the passionate committment that the belief is contained, as I argued in chapter one; to believe something is precisely to commit oneself to its truth; the arguments designed to abrogate scepticism only reinforce belief after the fact. What Kierkegaard is claiming is that we need to be honest with ourselves about the sources of our beliefs, that the epistemology of justification "rationally reconstructs" a fundamentally irrational process. It gives a rationalization for what we want or need to believe, and hence forces us to falsify our own doxastic lives. I think that that is as cogent as a criticism of contemporary as it is of traditional epistemology, and I will have more to say about why this is so in the next two chapters.


Again, after he goes through the process of the Meditations, Descartes professes to believe pretty much what he did before, but to believe it for much better reasons than he did before. But it is part of Kierkegaard's point that this is a rationalization in pejorative sense of the term, that Descartes knew exactly where he would end up before he began, and that the method of doubt becomes a method for disclaiming responsiblity for one's beliefs. They are not now merely beliefs; they have been objectively established. Certainly, for Kierkegaard, as far as the achievements of the epistemology of justification - man's attempt to lift from his shoulders the burden of his own subjectivity, to remove the emotive component from belief - is concerned, we are at best in doubt.


In Scepticism and Animal Faith Santayana goes so far as to assert that all knowledge is faith precisely because we are never in a position decisively to refute scepticism, that is, to remove objective uncertainty. Santayana sets out the familiar sceptical arguments elaborately, and in fact endorses their conclusions. That is, he affirms that we are in a position, as far as the exercise of reason with no assumptions is concerned, of radical and unrelievable doubt as to the existence of the external world, the deliverances of memory, even as to the existence or at least the nature of the subject. Or rather, if we were in fact creatures that generated beliefs by the emotionless exercise of reason with no assumptions, our doubts would be radical and unrelievable. But as agents, as passionate, individual creatures of the sort we are, we happen in fact to be under no serious doubt about these things. (The similarity of the position under consideration to the "naturalized" epistemology of Quine and others should here be remarked. Like Kierkegaard and Santayana, proponents of that project hold that scepticism is irrefutable and that how we actually come to acquire beliefs is relevant to how we ought to acquire them. And the similarity of the position to my reading of Diogenes, Johnson, and Moore need hardly be emphasized.) Santayana says: "the scepticism I am defending is not meant to be merely provisional; its just conclusions will remain fixed, to remind me perpetually that all alleged knowledge of matters of fact is faith only."(20)


This does not mean that all factual claims are wholly unjustified. Rather, Santayana's view is that all such claims rest finally on beliefs for which no justification can be produced. Knowledge, says Santayana, is faith mediated by signs. We take our intuitions to be signs of external objects and events. Taken in that way, such intuitions can lead to justified beliefs. But there is no justification for taking them that way. Animal faith, then, consists in the treatment of intuitions as signs of the external world, in "supposing that there is substantial there, something that will count and work in the world" (SAF p.39). The supposition that our intuitions are signs of external substances is the implicit and unjustified faith without which empirical knowledge would be impossible. Now in the next chapter I am going to argue that this picture of experience, the picture on which there is a certain internal component of experience (what Santayana terms `intuition') by which we move outward to the world, is radically misguided. But I suggest that it is still possible to reconstruct Santayana's argument here in a plausible way. That is, I do not think that it is possible to justify the completely general claims that there is an external world and that we have access to it. And even if it is possible to justify such claims, the view that we do not have any empirical knowledge until we have justified them is ludicrous. I will argue in the next chapter that we all know perfectly well that there are objects external to the mind.


In fact we are creatures, and not, for example, speculative philosophy itself, and an epistemology that hopes to elucidate our doxastic situation had better keep that fact firmly in mind. Hume's satisfaction as an agent as to the connection of cause and effect is all that he as an agent could hope to achieve, and he can achieve no other status than that of an agent. Philosophers, though they may seek to suspend or eliminate their own agency for the length of a book, retain their agency through the attempt. Every belief is someone's belief; every enduring belief consists of a relation (a relation of passionate commitment) of an agent to a proposition. For Kierkegaard the humanity and particularity philosophers retain is that of an inwardness, of the subjectivity of a spirit. For Santayana the humanity and particularity they retain is that of an individual organism, permeated with animal needs and endowed with specific faculties for their satisfaction. I certainly am more in sympathy with Santayana's naturalistic picture of doxastic agents. But what the views have in common epistemologically is equally clear: human beings are not the sort of things by or for whom objective certainty can be erected employing no assumptions; human belief is at base an emotional state. Now this is not at all to say that knowledge is impossible. But it provides a motivation for detaching knowledge from justification, a project the feasibility of which will be explored in the following chapters. But further, Santayana argues that since it is always possible to raise, and impossible decisively to answer, sceptical doubts, knowledge is always a form of faith.


Knowledge for such philosophers as Descartes and Hume seemed to require absolute certainty. And certainty here is again not merely a psychological matter; it is also logical. In order to be justified in one's belief, one must deduce it from self-evident or previously proven premises. On Santayana's view, knowledge is true belief based on experience; he calls that faith. But on the much weaker accounts of justification in the contemporary discussion, Santayana would hold an externalist theory of justification a la Goldman rather than the view that knowledge may include unjustified beliefs. Still, Santayana's discussion is relevant at least as an attack on the previous notion, and in fact is also useful in pointing out problems for certain contemporary views, notably foundationalism.



VIII. Philosophical Ingenuousness


Thus, universal doubt in Descartes' sense is for both Santayana and Kierkegaard impossible to entertain, and those who purport to be in doubt about certain claims are (as Hume so frankly admits in his own case) disingenuous, or else in a state of profound doxastic dissonance. Yet for Santayana and Kiekegaard, disingenuousness in philosophers is not lightly to be forgiven. One can claim to begin to philosophize without any presuppositions, but actually to begin without presuppositions, that is, without claims to which one's "merely logical intellect" has not been coerced, that is, to begin without faith: that is impossible. "Why cannot we remember to be human beings?" asks Kierkegaard (CUP p.103). That question could stand as a summary of Santayana's philosophy as well, though the two had radically different views about what it is to be a human being. As human beings, however, we cannot achieve anything approaching universal doubt; nor can we, as I have been and will be arguing, hold only justified beliefs. If we are told to begin with nothing, we may very well be puzzled as to how we are to begin at all. And to begin with nothing is not precisely to begin without anything, for the real beginning of this epistemological enterprise is the imperative that we ought to begin with nothing. That is itself a presupposition in the sense that it is only comprehensible as part of an entire epistemology. That we ought to begin with nothing: if this means something other than we ought not to begin, then it is itself part of a picture of ourselves and of philosophy to which our intellect is not coerced. That is to say, the method of doubt, if it is indeed to be regarded as a starting-point, itself rests on a certain variety of faith.


One way to view the faith on which methodological doubt and, in general, the epistemology of justification, rests is that it is faith in the truth: faith that we can find the truth and that we ought to try. I share this faith, but I simply deny that the procedures laid out in various theories of justification provide the only means by which the truth can legitimately be arrived at. This is the faith on which science, on a certain picture of science, depends. And indeed systematic doubt is associated with science, and its extension to all forms of inquiry and even as an ideal for the doxastic lives of particular persons is the symptom of scientism or science-worship (though it rests, I think, on a distorted view of how science in fact proceeds). The imperative that we ought to begin with nothing only makes sense if we believe not only that we can find the truth by the exercise of reason, but that the truth thus found is of paramount importance. This is a point that has been made forcibly by Nietszche, for example:


In science convictions have no rights of citizenship, as one says with good reason. Only when they decide to descend to the modesty of hypotheses, of a provisional experimental point of view, of a regulative fiction, they may be granted admission and even a certain value in the realm of knowledge - though always with the restriction that they remain under police supervision, under the police of mistrust. - But does this not mean, if you consider it more precisely, that a conviction may obtain admission to science only when it ceases to be a conviction? Would it not be the first step in the discipline of the scientific spirit that one would not permit oneself any more convictions?



Probably this is so; only we still have to ask: To make it possible for this discipline to begin, must there not be some prior conviction - even one that is so commanding and unconditional that it sacrifices all other convictions to itself? We see that science also rests on faith; there is simply no science "without presuppositions." The question whether truth is needed must not only have been affirmed in advance, but affirmed to such a degree that the principle, the faith, the conviction finds expression: "Nothing is needed more than truth, and in relation to it everything else has only second-rate value."(21)



And James makes exactly the same point:



Our belief in truth itself, for instance, that there is a truth, and that our minds and it are made for each other - what is it but passionate affirmation of desire, in which our social system backs us up? We want to have the truth; we want to believe that our experiments and studies and discussions must put us in a continually better and better position towards it; and on this line we agree to fight out our thinking lives. But if a Pyrrhonistic sceptic asks us how we know all this, can our logic find a reply? No! certainly it cannot. It is just one volition against another - we willing to go in for life upon a trust or assumption which he, for his part, does not care to make.(22)





Now there are several remarks I want to hang on these admirably clear passages. First, consider the hypothesis of philosophical scepticism, or universal doubt. For Descartes, this scepticism had to become more than a hypothesis; Descartes claimed in a certain way to have achieved doubt. One might say that Descartes, for example, doubted that there were objects external to the mind. Of course, he continued all the while to act as if the external world was not an illusion. Descartes himself, in the Discourse on Method admits that such facts might be relevant to an assessment of what one really belives:



[I]t seemed to me that to learn people's true opinions, I should pay attention to their conduct rather than to their words, not only because in our corrupt times there are few who are ready to say all that they believe, but also because many are not aware of their own beliefs, since the mental process of knowing a thing is . . . distinct from, and can occur without, the mental process of knowing that we know it.(23)







Now this is bizarre talk from a man who, earlier on the same page, writes: "reason obliged me to be irresolute in my beliefs, [but] there was no reason why I should be irresolute in my actions." Nevertheless, I suggest that Descartes is on the right track; our behavior, including our verbal behavior, is a good clue to where our actual commitments, that is, our actual beliefs, lie. I do not take a functional or behavioral approach to the theory of belief, as is evident from the first chapter. Nevertheless, such theories are surely correct when they emphasize that we can often infer what people do believe from their behavioral outputs. And I have arguing that our actual commitments are relevant to epistemological evaluation.


At any rate, James makes the same point: "The maximum of liveness in an hypothesis means willingness to act irrevocably. Practically, that means belief; but there us some believing tendency wherever there is willingness to act at all" (88). This is a felicitous way to put the matter. If we would like to know what a person believes, we do well to consult her actions. To be a hypocrite is not to believe something and to act incompatibly with that belief; it is to profess a belief one does not hold, the fact that one does not hold it being demonstrated by one's actions. Belief, that is, is a relation, a passionate relation, as James insists, of a particular person to a particular claim (via a representation); one does not believe or fail to believe something merely on paper. As Nietzsche puts it: "It makes the most telling difference whether a thinker has a personal relationship to his problems and finds in them his destiny, his distress, and his greatest happines, or an `impersonal' one, meaning that he can do no better than to touch them and grasp them with the antennae of cold, curious thought" (283). The relationship is indeed telling; I would hold (with James) that it is in fact the difference between belief and a certain sort of non-belief, that is, non-belief in which belief is professed. That is because belief from which the emotive component has been subtracted is not belief at all.


Now when science sets up hypotheses, the scientist witholds his assent and maintains a healthy doubt. At least, that is the ideal within a certain vision of the scientific project. (As James says, it is far more likely that the scientist sets out to confirm a claim to which he is already passionately committed.) But what James and Nietzsche assert, and I should think this could hardly be denied, is that underlying this doubt, or this exquisite suspension between belief and disbelief which turns a claim into an hypothesis, there is a deeper faith, a faith that there is a truth of the matter, that it is possible and also desirable to reach it. This faith will be largely dispositional; it is so deeply buried in the foundations of science (as well as of philosophy), that it is difficult to bring it to consciousness. Indeed, it seems vaguely ludicrous to do so. And I am by no means arguing that we ought to leave this faith behind; I am simply arguing that we ought to acknowledge that we do have this faith. Like the faith that the external world is not an illusion, it is needed for science to continue. Thus, the doctrine of scientism - that science is the model for the acquisition of knowledge, and that it consists of an exercise in reason purged of emotion - is false and impossible. There is, in Nietzsche's terms, no science without presuppositions, without passionate commitments.


But it is Nietzsche's point that even if we were to accept the scientist as a model of how we ought to conduct ourselves doxastically, we would still not be able to expunge faith. The faith of the scientist consists of precisely those assumptions which allow science to arise and which keep it in operation. So it should in a certain sense be no surprise that the continuing success of science seems to "confirm" that fundamental faith. But it is certainly odd for the champions of "reason" in this scientistic sense to revile other forms of faith, other unjustified beliefs (paradigmatically, for James and Kierkegaard, religious faith) when they, in James' words, are chock-full of some faith or other themselves. Nietzsche:



It is no different with the faith with which so many materialistic natural scientists rest content nowadays, the faith in a world that is supposed to have its equivalent and its measure in human thought and human valuations - a "world of truth" that we can master completely and forever with the aid of our square little reason. What? Do we really want to permit existence to be degraded for us like this - reduced to a mere exercise for a calculator and an indoor diversion for mathematicians? Above all, one should not wish to divest existence of its rich ambiguity: that is a dictate of good taste, gentlemen, the taste of reverence for everything that lies beyond your horizon. That the only justifiable interpretation of the world should be one in which you are justified because you can continue to work and do research scientifically in your sense (you really mean, mechanistically?) - an interpretation that permits counting, calculating, weighing, seeing, and touching, and nothing more - that is crudity and naivete, assuming it is not a mental illness, an idiocy. (335)



Now James by no means regarded science as a mental illness, nor did he give any reason to to jettison logic or epistemological inquiry. But he nevertheless affirmed that those who attack faith themselves take what faith they need, and then try to deny to others the faith they need: "This very law which the logicians would impose upon us - if I may give the name of logicians to those who would rule out our willing nature here - is based on nothing but their own natural wish to exclude all elements for which they, in their professional quality of logicians, can find no use" (94). It must be remarked that, when these people are no longer acting "in their professional quality," that is, taking up the "external" view, they, like everyone else, believe what they must believe to live.


Here, James' distinction between two epistemological imperatives is relevant. He distinguishes between the desire to know the truth and the desire to avoid error. These, as he points out, are by no means two ways of stating the same goal; the former might invite us to believe what we cannot demonstrate to be the truth, in order to give ourselves some chance at actually arriving at the truth, while the latter would have us suspend judgement until a decisive argument can be produced. James sets out this point in one of the most disarming and characteristic passages in all his writings:



We must remember that these feelings of our duty about either truth or error are in any case only expressions of our passional life. Biologically considered, our minds are as ready to grind out falsehood as veracity, and he who says, "Better go without belief forever than believe a lie!" merely shows his own preponderant horror of becoming a dupe. . . . For my own part, I have also a horror of being duped; but I can believe that worse things than being duped may happen to a man in this world. . . . Our errors are surely not such awfully solemn things. In a world where we are so certain to incur them in spite of all our caution, a certain lightness of heart seems healthier than excessive nervousness on their behalf. (100)







To avoid error, the sceptic witholds assent. But the epistemological imperative remains in the background as an unstated but obviously important assumption. Nor can this assumption stand aloofly off as the sceptic's only faith. For the very notions of truth and error presuppose the whole fundamental machinery of logic, of the principle of non-contradiction and its elaborations and kindred principles.


I will discuss certain contemporary accounts of knowledge that require justification in the next two chapters. But to anticipate here, I think these accounts, vastly divergent though they are, are infected by precisely this sort of scientism. They hold that as we go about generating beliefs in the practical, everday world, we ought to be operating like an (idealized) scientist: suspending beliefs until the evidence is in, testing hypotheses to which we have no commitment, and so forth. But it is a contribution of the pragmatists as well as the existentialists to admit frankly that this is at the least extremely difficult in the situation in which we find ourselves, and also undesirable. I subscribe to Santayana's view that human beings are merely one species of animal, and that we are under the epistemological contraints that that fact entails: we are obliged to act, and we are obliged to believe. The ultimate epistemological injunction should not be to believe only what one is justified in believing, but to declare oneself to believe what one does in fact believe, and to examine those beliefs that can be questioned.


If we are to begin at all, then, in a philosophical or scientific inquiry, even if we end with a suspension of judgement, we are going to have to begin in a state of faith, in a state of credulity with regard to certain claims unjustified by argumentation. Otherwise, the inquiry which leads us either to hold certain beliefs or to withold all beliefs will never begin. But now both James and Nietzsche ask: What is the force of the claims we must accept, what is their origin, their character? What lends them the power they have over us? For example, what kind of claim is it that we ought to believe all and only what we are justified in believing?




IX. The Ethics of Belief


Oddly enough, for both James and Nietzsche, the claim that we ought to believe all and only what we are justified in believing is a moral claim. And I think this is plausible. The epistemology of justification, in both its traditional and its contemporary guises, depends on a certain moral injunction concerning honesty. James makes this point by quoting proponents of the view. Here is a passage James quotes from Thomas Huxley: "My only consolation lies in the reflection that, however bad our posterity may become, so far as they hold by the plain rule of not pretending to believe what they have no reason to believe, because it may be to their advantage so to pretend, they will not have reached the lowest depth of immorality" (93). Now as I hope we have seen, it is perhaps less easy than Huxley imagined to believe only what we have good reason to believe, and to eliminate the element of desire from belief. But what is relevant here is the sense of moral outrage, the feeling that faith is not, or is not only, an intellectual failing, but an ethical one. And James quotes William Clifford to similar effect: "If a belief has been acccepted on insufficient evidence [even if it is true] the pleasure is a stolen one. . . . It is sinful because it is stolen in defiance of our duty to mankind."


James offers the following reply:



Science herself consults the heart when she lays it down that the infinite ascertainment of fact and correction of false belief are the supreme goods for man. Challenge that statement, and science can only repeat it oracularly, or else prove it by showing that such ascertainment and correction brings man all sorts of other goods which man's heart in turn declares. . . . If your heart does not want a world of moral reality, your head will surely never make you believe in one. . . . Moral scepticism can no more be refuted or proved by logic than intellectual scepticism can. When we stick to it that there is truth (be it of either kind), we do so with our whole nature, and resolve to stand or fall by the results. (103, 4)



To repeat, to believe without reasoned justification is supposed by those whom James refers to as "logicians," and to whom I have been referring as proponents of the epistemology of justification, to be "sinful."


This is a particularly apt turn of phrase with regard to Nietzsche's formulation of the problem, wherein the faith of the epistemology of justification is a particular faith, namely, and surprisingly enough, Christian faith. Now James proposed that science can only justify its insistence on justification by pointing to its utility. Nietzsche, on the other hand, claims that the justification of belief and the quest for certainty (in Dewey's famous phrase) do not have the sort of utility that is often claimed for them. And indeed this is the case, as perhaps someone like Clifford would be happy to admit. It can be useful to believe a truth, on occasion, and just as useful on other occasions to believe a falsehood. Nietzsche:



What do you know in advance of the character of existence to be able to decide whether the greater advantage is on the side of the unconditionally mistrustful or of the unconditionally trusting? But if both should be required, much trust as well as much mistrust, from where would science then be permitted to take its unconditional faith or conviction on which it rests, that truth is more important than any other thing, including every other conviction? Precisely this conviction could never have come into being if both truth and untruth constantly proved to be useful, which is the case. Thus - the faith in science, which after all exists undeniably, cannot owe its origin to such a calculus of utility; it must have arisen in spite of the fact that the disutility and dangerousness of "the will to truth," of "truth at any price" is proved to it constantly. . . . Consequently, "will to truth" does not mean "I will not allow myself to be deceived" but - there is no alternative - "I will not deceive, not even myself"; and with that we stand on moral ground. (281, 2)





Thus, the fundamental faith of "science," or "the logician" is not epistemological at all, but moral. And further, according to Nietzsche, it is Christian morality (which he also describes as the faith of Plato), the morality of an other-worldly realm, the morality the source of which declares "I am the way, the truth," that informs the epistemolgy of mistrust, of the quest for certainty.


But conceived of as the injunction that one should not deceive oneself, the epistemology of justification is in one sense an abject failure. That is, it can easily lead precisely to a systematic self-deception; it can lead to the view that one believes only what it is reasonable to believe and not at all what one would like to believe. But as I have suggested, such a goal is much harder to achieve than it might appear. In short and again, it is extremely difficult to entertain a thoroughgoing scepticism. The fact that some philosophers claim to do so seems to Santayana a moral failing, a piece of dishonesty and a denial of their own nature. So the moral danger lurks both ways: in too much and in too little self-examination. To Kierkegaard the whole thing appears to be rather bathetic. He says: "All scepticism is a kind of idealism. . . . There is no special difficulty connected with being an idealist in imagination; but to exist as an idealist is an extremely strenuous task, because existence itself constitutes a hindrance and an objection. . . . to understand everything except one's own self is very comical" (CUP pp.315,16).


And Santayana makes the same point even more forcefully:



I for one will honour [the sceptic] in his sincerity and in his solitude. But I will not honour him, nor think him a philosopher, if he is a sceptic only histrionically, in the wretched controversies of the schools, and believes in substance again when off the stage. I am not concerned with make-believe philosophies, but about my actual beliefs. It is only out of his own mouth, or rather out of his own heart, that I should care to convince the sceptic. Scepticism, if it could be sincere, would be the best of philosophies. But I suspect that other sceptics, as well as I, always believe in substance, and that their denial of it is sheer sophistry and the weaving of verbal arguments in which their own most familiar and massive convictions are ignored. (SAF 186)





The proposition "Here's dinner" entails that there are objects external to the mind. Hence, a person who sincerely utters that proposition has committed himself to the claim that there are objects external to the mind, and also to such claims as that he is not currently dreaming, that an evil genius is not currently deceiving him, and that he is not a brain in a vat. So a philosopher who continues to produce such utterances while claiming on paper to doubt that there are objects external to the mind is a hypocrite.


X. Faith and Reason


For Santayana, the position of the Cartesian epistemologist or speculative philosopher is problematic, first of all, because it is unreasonable. This claim is of course designed to enrage those against whom it is made, who regard themselves as the champions of reason. But for Santayana, if any person were in fact able to get himself into the position that the Descartes regards as a starting point, that person would have made reason itself impossible for himself. Reason according to Santayana acts as a mechanism that jettisons superstition, that discards what we do not need, but only desire to believe. It performs intellectual hygeine; it gets rid of all but the minimum core of beliefs and gives the rest the status of hypotheses, sets them up for examination: "the things posited by animal faith in the heat of action [are] the only things in which there is any reason for believing" (SAF p.232).


Reason could not cause us to discard all belief whatever, because we are literally incapable of doubting some claims, and because reason would have to destroy itself in order to perform that task:



It might seem ignominious to believe something on compulsion, because I can't help believing it; when reason awakes in a man it asks for reasons for everything. Yet this demand is unreasonable: there cannot be a reason for everything. It is mere automatic habit in the philosopher to make this demand, as it is in the common man not to make it. When once I have admitted the facts of nature, and taken for granted the character of animal life, and the incarnaton of spirit in this animal life, then indeed many excellent reasons for the belief in substance will appear; and not only many reasons for using the category of substance, and positing substance of some vague, ambient sort, but reasons for believing in a substance rather elaborately defined and scientifically describable in many of its habits and properties. But I am not yet ready for that. Lest that investigation, when undertaken, should ignore its foundations, or be impatient of its limits, I must insist here that trust in knowledge, and belief in anything to know, are merely instinctive and, in a manner, pathological. (SAF p.187)





Santayana declares that we should frankly acknowledge the doxastic situation in which we really do find ourselves, and that reason is of service only within that actual situation. It is just this point which motivates the argument in the last chapter of this book that knowledge is merely true belief. That is, I regard the fundamental test of theories of knowledge as being whether they account for the antecedent facts about what we do know. Now it might be asserted that, in doing epistemology, we ought to make no assumptions about what we know; specification of the facts about what we know ought to fall out of a theory of knowledge, rather than vice versa. But this picture itself begs the question in favor of the epistemology of justification; it demands that we "strip ourselves of mere assumptions" in the classic Cartesian sense. The problem is, as I have been arguing, that when we attempt to fo this, we sink into doxastic dissonance, or, more often, into sheer hypocrisy.


To demand that we follow an impossible procedure for the acquisition of knowledge, for example, to make it a requirement on knowledge that it be established by a procedure that is known to be reliable, must leave us in scepticism or lead us finally to validate faith of one kind or another. At present, faith in James' sense and about one thing or another is extraordinarily difficult to avoid. We are not free to suspend belief on such questions as whether bread will nourish us or walls protect us; our practice does not commit us to such claims, it shows that we are already committed to them. That is what Santayana designates animal faith; were we not animals in need of nourishment and protection from the elements, these particular beliefs would be dispensible, or rather, would never have arisen. That they are indispensible is not of course a demonstration that they are true (though they are for the most part true), but we should die of starvation if we were to await a demonstration, that is, if we did not first perish of exposure: "whilst all the animals trust their senses and live, philosophy would persuade man not to trust them and, if he was consistent, to stop living" (SAF p.297).


Santayana, in Reason in Religion, defends what he terms `piety'; though this piety is not religious in the sense that its object is the divine:



Piety, in spite of its allegories, contains a much greater wisdom than half-enlightened and pert intellect can attain. Natural beings have natural obligations, and the value of things for them is qualified by distance and by accidental material connections. Intellect would tend to gauge things impersonally by their intrinsic values, since intellect itself is a sort of disembodied and universal function; it would tend to disregard material conditions and that irrational substratum of reason without which reason would have no organs and no points of application. Piety, on the contrary, esteems things apart from their intrinsic worth, on account of their relation to the agent's person and fortune. Yet such esteem is perfectly rational, partiality in man's affections and allegiances being justified by the partial nature and local status of his life. Piety is the spirit's acknowledgement of its incarnation.(24)





This is an affirmation of the fact that we are passionately concerned with the world, and that this is why we have beliefs in the first place.


For Kierkegaard, also, to acknowledge faith is to acknowledge our status as existing subjects, whereas blithely to claim that one has doubted everything and then continued is to deny this status:



In our time the processs of abstracting from everything has been relegated to the printed page, just as the task of doubting everything is disposed of once and for all on paper. One of the things that has given rise to so much confusion in modern philosophy is that philosophers have so many brief sayings about infinite tasks, and respect this paper money among themselves, while it almost never occurs to anyone to try to realize the posited task. In this way everything is easily finished, and it becomes possible to begin without presuppositions. The presupposition of a universal doubt, for example, would require an entire human life; now, it is no sooner said than done. (CUP P.282)





This is something which not only philosophers such as Descartes and Hume, but also participants in the contemporary debates in epistemology, would do well to heed. When we ask the question of what it would be a good thing for us to believe, when we ask what we know, we must keep in mind that the agents about whom we are asking such questions are ourselves, are particular human beings, particular animals. As William Lycan puts it, "knowledge" is our word, and it has certain well-recognized applications. If we begin to discover that, on the account we are giving of knowledge, it does not appear certain that anyone knows anything, or more specifically that it is not the case that everyone knows that here is a hand in the circumstances provided by Moore, whether or not he has a justification available, we have left our words and ourselves behind.


So I want to read Kierkegaard, Santayana and James as doing systematically and elaborately what Diogenes, Johnson, and Moore do, as it were, anecdotally, that is, showing us where our genuine commitments lie, and in the process elucidating who we are. Like Diogenes, Johnson, and Moore, neither Kierkegaard nor Santayana attempt to refute scepticism. Rather, they pre-empt it by relating or reducing abstract "doubts" to acts (or events) of doubting. And this is the only construal of doubt that is compatible with the claim, made in the last chapter, that doubt is an attitude toward a proposition taken up by a particular doxastic agent. By the same token, Kierkegaard, James, and Santayana do not attempt to justify faith. Rather, they describe it, and relate or reduce abstract "beliefs" to acts (or events) of believing, acts with real emotional content. Where there are "doubts," there someone doubts, and the application of ingenuity may or may not extricate him. Where there are beliefs, there someone believes, and here the application of ingenuity may even, in some cases, be trifling or useless.



At the beginning of this chapter, I argued that, if externalism (or, more precisely, reliabilism) is an admissable strategy for justification, then, if justification is required for knowledge, empirical knowledge is, at a minimum, possible. But I also argued that the sceptical problem recurs at a second level; that, even if we could come to know propositions about the external world by a reliable mechanism, we would not be in a position to establish when we are dreaming and when we were not, and thus that it would seem rational to suspend judgement about all empirical propositions. But I argued that we ought not to suspend judgement after all, and hence that we do not need to establish, in a general way, that some empirical beliefs can be justified. It is time, however, to turn to the full-fledged internalist view of Descartes and contemporary foundationalists such as Chisholm. I will argue, then, in the next chapter, that the internalist conception of justification pre-supposes a false view about human empirical experience in general.


1. Rene Descartes, Meditations on First Philosophy, trans. Donald Cress (Indianapolis: Hackett, 1979), p.13.


2. Diogenes Laertius, Lives of Eminent Philosophers, trans. R.D. Hicks (London: William Heinemann, 1925), VI. 39.


3. James Boswell, Life of Johnson (Oxford: Oxford University Press 1985), p.333. As Boswell suggests, Johnson's refutation is hopeless against Berkeley. I will not consider it as a refutation of idealism, but as an answer to scepticism about the external world.


4. G.E. Moore, "Proof of an External World," Philosophical Papers (London, George Allen and Unwin, 1959), pp.145,6.


5. A similar point is made by Paul Moser in his review of Stroud's book in The British Journal for the Philosophy of Science 37 (1986) 235-8.


6. See the seventh Meditation and the third Dialogue Between Hylas and Philonous.


7. William Alston, "Level Confusions in Epistemology," Epistemic Justification (Ithaca: Cornell University Press, 1989), pp.153-171.


8. It is sometimes thought that the problems raised, e.g., by the dream argument concern only knowledge, and do not, if accepted, rationally require us to suspend judgement in those matters of which the argument purports to undermine our knowledge. This approach, to repeat, cannot be sustained. If I really might be dreaming, and really cannot show that I am not, then why as a rational agent ought I to maintain empirical beliefs? The point is made succinctly by M.F. Burnyeat: "if the evidence of our senses is really shown to be unreliable and the inferences we ordinarily base on this evidence are unwarranted, the correct moral to draw is not merely that we should not claim to know things on these grounds but that we should not believe them either." "Can the Skeptic Live His Skepticism?" The Skeptical Tradition (Berkeley: University of California Press, 1983), p.119.


9. This has been denied by, e.g., Merrill Ring, in "Knowledge: the Cessation of Belief," American Philosophical Quarterly, vol. 14 (1977), pp.51-59. I find this position perverse. Not only do I know that the sky is blue, I firmly believe it.


10. I owe this point to Paul Humphreys.


11. M.F. Burnyeat gives a similar interpretation of the initial demonstration in "Examples in Epistemology: Socrates, Theaetetus, and G.E. Moore," Philosophy, 52 (1977), p.395. Burnyeat goes on to trace the (from my point of view unfortunate) development of Moore's thought.


12. Moore, "Some Judgements of Perception," Philosophical Studies (New York: Harcourt, Brace & Co. 1922), p.228. The point is emphasized in Charles Larmore's review of Stroud in The Journal of Philosophy, 84 (1987) 384-92.


13. Marie McGinn, Sense and Certainty (Oxford: Basil Blackwell, 1989), p.46.


14. I do not, however, disagree with Burnyeat's conclusion in "Can the Skeptic Live His Skepticism?" that thoroughgoing scepticism, scepticism not limited, e.g., to claims about the external world, cannot be lived.


15. See, e.g. Concluding Unscientific Postscript (CUP), trans. David F. Swenson and Walter Lowrie (Princeton: Princeton University Press, 1968), p.286.


16. William James, "The Will to Believe," Essays in Pragmatism (New York: Hafner, 1948), p.88.


17. Soren Kierkegaard, Johannes Climacus, or De Omnibus Dubitandum Est, trans. T.H. Croxall (Stanford: Stanford University Press, 1958), p.146.


18. Soren Kierkegaard, Fear and Trembling, trans. Walter Lowrie (New York: Doubleday: 1954), p.22.


19. David Hume, An Enquiry Concerning Human Understanding (Indianapolis: Hackett, 1977), p.24.


20. George Santayana, Scepticism and Animal Faith (SAF) (New York: Dover, 1955), p.49.


21. Friedrich Nietzsche, The Gay Science, trans. Walter Kaufmann (New York: Vintage, 1974), pp.280, 1. Further references will appear in the text by page number.


22. James, "The Will to Believe," p.94. Further references will appear in the text.


23. Descartes, Discourse on Method, trans. Laurence J. Lafleur (Indianapolis; Bobbs-Merrill, 1960), pp.18,19.


24. Santayana, Reason in Religion (New York: Charles Scribner's Sons, 1936), pp.183, 4.






Chapter III



Radical Externalism With Regard to Experience





I. Internalism



This chapter is an attempt to construct an alternative to a position I will term "internalism with regard to veridical experience," and to use a rejection of this view to motivate a rejection of the epistemology of justification. Internalism with regard to veridical experience is often associated with epistemological foundationalism, but it has analogues or versions in other or related traditions, for example classical empiricism, positivism, and phenomenalism. Indeed, even some proponents of externalist theories of epistemic justification hold a form of internalism with regard to veridical experience.(1) At any rate, the position is this: veridical empirical experience, experience of the world, can be usefully bifurcated into inner and outer components. Experience of the world, on the view in question, has a certain phenomenological character, a certain subjective side which is intrinsic or internal to the experiencer and hence can be considered in isolation from the external-world situations which it may represent or to which it may give us access. In addition, this internal component of veridical experience is epistemically available to the experiencer; it is a source of beliefs or knowledge about the external world. That this alleged inner aspect or component of veridical empirical experience can be considered in isolation is suggested by cases in which apparently empirical experience fails to give us access to external world situations: experiences of dreams, illusions, hallucinations, and so forth. Such cases seem to give us the inner without the outer component of empirical experience, thus suggesting (or so it is held) that the two can be separated also in the case of veridical experience.


Some versions of internalism with regard to veridical experience refer to the epistemically available inner component of experience as `sense data,' `ideas,' `sensations,' `appearances,' `representations,' `seemings.' But I will not be concerned with general characterizations of this component; I will not be explicitly concerned with the question of whether, for example, there are such things as sense data, though the view that I sketch will entail that there are no sense data in play in ordinary veridical experience of the world. It is rather to the general division of experience on which these views rely that I will address my remarks.


On typical versions of this view (and here we move from philosophy of mind to epistemology), we infer that certain external world situations obtain from the phenomenological character of experience; we move outward through the inner component of experience to real world situations. Again, I want to place no particular weight on the notion of inference here; on certain versions of the view, the inner component of experience need never be formulated into beliefs of the sort that can serve as parts of an inference. John Pollock, for example, says "We generally `look through' our sensations at the world."(2)


This need not be construed as inferential, but such a view will be equally under attack here.


So, putting it at its most general, internalism with regard to veridical experience is embodied in three theses:





(1) Veridical experience can be divided into inner and outer components.


(2) Some aspects of the inner component of veridical experience are epistemically available to the experiencer.


(3) The experiencer moves outward epistemically through the inner into the outer component, that is, comes to know or believe claims about the external world by undergoing the inner component of experience.



I think that (2) and (3) are false, and that they lead to a highly distorted epistemology.


These three theses are very widely held indeed. Nevertheless, it will be useful to summarize a contemporary epistemological theory that explicitly endorses them. I will start, then, with Roderick Chisholm's view as expressed in "A Version of Foundationalism." Chisholm describes certain properties as being "self-presenting." "If the property of being F is self-presenting," he writes, "then for every [person] x, if (1) x has the property of being F, and if (2) x considers his being F, then it is certain for x that he has the property of being F."(3) For example, if I am sad, and I consider whether I am sad, it will be certain for me that I am sad. And, Chisholm adds with regard to empirical experience, if I am having the experience of red, if I seem to see red or am, in Chisholm's adverbial formulation, appeared to redly, then if I consider the matter, it will be certain for me that I am having such an experience or am being so appeared to. He writes:



It may be self-presenting for x that he is being appeared to in a certain way (that he senses in a certain way). But it cannot be self-presenting to him that there is something that is appearing to him in that way (i.e. it cannot be self-presenting to him that an external stimulus object causes him to sense in that way). By means of what principle, then, can the person pass from a way of appearing to a particular physical thing that "transcends" that way of appearing? (p. 302)



The inner and outer components of experience here are distinguished by the supposed fact that whereas the inner component of experience may be self-presenting, the outer component of experience never is. Chisholm ends up supplying a principle of the kind he describes above, a principle that allows us to move from the inner to the outer component of experience, and some such principle is required on every epistemological version of internalism which does not lead directly to external-world scepticism. Chisholm claims that being appeared to in certain ways may makes it evident to x that he is experiencing a certain external object. `Making it evident' is not, for Chisholm, exactly a matter of drawing inferences, but the difference need not concern us here, though I will have something to say about it later in this chapter.



II. Fusion


Initially, I will not attempt to refute or even directly attack internalism with regard to veridical experience, though I will eventualy give some considerations which count against it. To begin with, I will attempt merely to formulate an alternative view (a task which turns out to be considerably less straightforward than it might seem), and to answer certain traditional arguments that might be mustered against it.


At any rate, I can have the experience of seeing (to use an example from Austin) a pig, and no doubt in this experience I come to display certain properties. But I will argue that the properties of me that are epistemically relevant are not isolable from the state of affairs in which the experience comes to be produced, namely a particular juxtaposition of me with a pig, because they are properties I display only in relation to, only within, this state of affairs. Now many thinkers, notably Hilary Putnam, Tyler Burge, and Colin McGinn, have argued that mental states which concern external objects cannot be individuated except with reference to facts about the external world that are not necessarily within the ken of the cognizer.(4)


I think that is true. But here I am pursuing a rather more radical externalist claim, one which these thinkers might well reject.(5) I shall suggest that, in a case where one is experiencing an external object, there are no intrinsic or internal properties of the experiencer whatsoever which she displays in virtue of experiencing that object, of a sort which are necessary for her to have that experience, and which are epistemically accessible. To repeat, my aim at the moment is merely to give a clear statement of the view, and to explain some of its epistemological implications.


Now it is sometimes held that some sort of externalism with regard to experience is supported by certain familiar facts about the meaning of the word `experience.' Note that I cannot have the experience of a pig if a pig is not actually present, just as I cannot see or hear a pig if there is no pig there. In addition, experience, in contexts such as "S has an experience of x" introduces an extensional or transparent context. Ascriptions of experience of this kind are open to substitution of co-referring expressions. If the pig is called the `Empress of Blandings,' and I have an experience of the pig, I have an experience of the Empress of Blandings, whether or not I know that `Empress of Blandings' is a name of this pig. This shows at a minimum that `experience' is not always used to ascribe a purely internal state, that in certain contexts, at any rate, `experience' is meant to pick out a relation between a cognizer and objects or states of affairs in the world. For if this sort of experience were a matter of my internal representations, substitution salva veritate of co-referring expressions would not always go through; it would go through only for expressions I took to refer to the pig in question.


Such facts, however, do not even suggest that internalism is false. An internalist who holds (1)-(3) will say this: the pig causes my visual apparatus to assume a certain intrinsic state. I then infer from beliefs about this state that I am in the presence of a pig. (Alternately: my intrinsic state makes it evident to me that I am in the presence of a pig.) Of course, we can say truly that I experience a pig only in the case where it is pig which causes my visual apparatus (including whatever parts of the brain are employed in vision) to assume this state. And of course, in that case, an ascription of the experience will be extensional. So at the stage at which a division is drawn for epistemic purposes between inner and outer components of experience, the internalist has no difficulty whatever in accounting for these facts about linguistic usage. But it nevertheless seems to me that there are no intrinsic features of the kind described; the sort of properties of me that are necessary for me to see a pig and that are epistemically available to me in normal circumstances cannot be isolated in the required way; they are not intrinsic or internal properties of me.


This is not as mysterious as it may sound. We can clarify it by analogy. To ascribe to S an experience of x is to ascribe a relation that holds between S and x, or, alternatively, to ascribe a certain relational property to S. Now consider the relational property I ascribe to S when I say: `S is located to the south of Manchester.' This does ascribe a certain property to S: it ascribes to S the property of being located to the south of Manchester. But this property is not intrinsic to S. More precisely, it is not necessary that, in order to be located to the south of Manchester, S display any particular intrinsic properties in virtue of standing in that relation. The state which S instantiates she instantiates in virtue of certain external-world relations. If there were any "internalists" with regard to such properties, they would be obliged to claim that, in virtue of being located to the south of Manchester, there are certain intrinsic properties that S displays of a sort that are necessary for her to be so located. But there are no intrinsic properties of S which she displays in virtue of being located south of Manchester and the display of which is also necessary for her to be so located. However, there are intrinsic properties of S without which she could not be located to the south of Manchester, such as being spatially circumscribed. And there are (possibly), as we shall see, intrinsic properties which she displays in virtue of being located south of Manchester. But no property of S meets both these conditions.


Now it will surely be objected that, unlike the relation that S displays to Manchester when she is located south of Manchester, the relation S displays to a pig when she experiences a pig is a causal relation. In an experience of a pig, S is operated on causally by a pig. That is true. But it does not follow that there are intrinsic properties of S that she displays in virtue of being causally operated on by a pig, and the display of which are necessary conditions of her standing in that relation. That is to say, it does not follow that there are properties that S displays in virtue of being causally operated on by a pig that are isolable from the external world situation in which she comes to be so causally operated on. Consider the transmission of motion among billiard balls. The cue ball strikes the object ball, causing it to carom off the bumper, let us say. In this causal situation, the object ball comes to be in motion. But of course, we cannot characterize the motion of the object ball if we restrict our attention to the intrinsic or internal properties of the ball. Motion is a relative property, a property of something in a situation.


Let us, then, introduce some terminology:



(F) A relation F between two individuals a and b is fused with respect to b just if there are no intrinsic properties that b displays in virtue of satisfying Fax which are necessary in order for b to satisfy Fax.





It is time to explain this last clause. It is required in order to block the following sort of counter-example to the fusion of spatial location. Emily is located to the south of Manchester at 6:12 p.m.. At 6:12 p.m. it rains south of Manchester, and Emily gets wet. Being wet is (more or less) an intrinsic property of Emily. Thus there is an intrinsic property that Emily displays in virtue of being (because she is) located to the south of Manchester. But she might not have been wet at 6:12, for example, if it did not rain south of Manchester at that time; being wet is not necessary for her to be so located. So in order to capture the fact that the relation of being located to the south of Manchester is fused with respect to Emily, we must include this final clause in (F). The clause is relevant to the treatment of experience, for those that hold (1), (2), and (3) are likely to hold that we could not have experiences of the external world without undergoing the inner component of such experiences, however the latter is characterized. So at least the kind of properties displayed will be necessary for the `experience' relation. That is how I construe the final clause of (F), as ranging over kinds of properties.


Both the relation of being located to the south of Manchester and the relation of being set in motion by the cue ball are fused in the sense specified in (F). But we need to further distinguish the two cases:



(F1) F is partially fused with respect to b just if there are no properties that b displays in virtue of satisfying Fax which can be specified without reference to some feature of the world external to b, and which are necessary for b to satisfy Fax.



Consider again the billiard ball example. Here we can specify the motion of the object ball without reference to the motion of the cue ball, though not without reference to features of the world external to the object ball. We need a stronger formulation, however, to deal with, anong others, relations of spatial location:



(F2) F is fully fused with respect to b just if there are no properties that b displays in virtue of satisfying Fax which can be specified without reference to a, and which are necessary for b to satisfy Fax.



The relation of being located to the south of Manchester is fully fused. We cannot specify the properties of S which she displays in virtue of being located to the south of Manchester (and which are of a kind necessary for her to be so located) without referring to Manchester. It might follow from the fact that S is located to the south of London that she is located to the south of Manchester. But this follows only with the added premise that London itself is located to the south of Manchester.


Now the claim that experience is partially, much less fully fused, is implausible. Whatever the details of a good physiological theory of vision, it is obvious that seeing something causes a series of events in the eye and in the brain. And it is equally obvious that some of these events can be specified without appeal to the world outside the head of the experiencer. For example, certain neurons may fire, and the firing of neurons is internal to the experiencer. In addition, the firing of these neurons, or at any rate of some neurons, and so forth, may well be necessary conditions, at least physiologically necessary conditions, of having the experience in question.


These facts, however, are not yet enough to comfort a foundationalist who holds both (2) and (3). For according to (2), the relevant internal aspects of the experience must be epistemically available to the experiencer. And of course, this is just what the internalist asserts. But the firing of neurons and so forth is not, or is at least not in general, epistemically available to the experiencer. In a normal case, we do not infer from the firing of neurons that we are experiencing a pig; indeed in the normal case we have no beliefs about the firing of neurons in such a case; we have no access to the firing of our neurons whatever. (Or, at any rate, we have no access to the firing of neurons under that description.) And even if we did, it would be perverse to move outward epistemically from beliefs about the firing of neurons to beliefs about the external world; the beliefs about neurons would seem to be otiose.


So the internalist, to repeat, needs to assert that the inner aspect of experience is epistemically available. Let us consider a more fine-grained way to characterize fusion in relations.



(FA1) A relation F between two individuals a and b is partially fused with respect to b under aspect G just if G is a property b displays in virtue of satisfying Fax, the display of which is necessary in order for b to satisfy Fax, and G cannot be specified without reference to some feature(s) of the world external to b.



(FA2) A relation F between two individuals a and b is fully fused with respect to b under aspect G just if G is a property that b displays in virtue of satisfying Fax, which is necessary in order for b to satisfy Fax, and G cannot be specified without reference to a.



We are now in a position to formulate two versions of externalism with regard to empirical veridical experience. Let us term the first simply "externalism with regard to veridical experience":



(E) All properties of a person which she displays in virtue of experiencing an external object, and which are epistemically available to her with reference to the existence and properties of that object, are aspects under which her experiential relation to that object is partially fused.



This is a fairly mild principle, though still no doubt controversial: it says that veridical experience presupposes an external world, and that the epistemically available aspects of that experience cannot be specified without reference to some features of that world.


The view I have been attempting to formulate, however, might be termed "radical externalism with regard to veridical experience":



(ER) All properties of a person which she displays in virtue of experiencing an external object, and which are epistemically available to her with reference to the existence and properties of that object, are aspects under which her experiential relation to that object is fully fused.



To put it less technically, though somewhat misleadingly, there is no epistemically accessible internal component of veridical empirical experience. Among other things, there are no sense data, phenomena, or seemings involved in veridical empirical experience.


It may well be asserted that (ER) leads directly to external-world scepticism. If I do not have epistemic access to the internal component of my veridical experience, how can I know anything at all about the external world? The proponent of radical externalism may reply as follows: In my experience of a pig, aspects of me are fused with a pig. (I admit this sounds a bit disquieting.) I take the role of b in schema (FA2). In such circumstances, I (may) have epistemic access to the pig, unmediated by any internal aspects of me. (To repeat, there are internal aspects of me which are necessary for me to experience a pig, and which I display in virtue of seeing a pig. But these are not (in a normal case) epistemically available.) So (ER) is going to lead rather naturally to some form of direct realism.


And since I have already argued at length that we need not take scepticism about the external world seriously, we do not need to take it seriously here.



III. The Argument From Illusion


Now if (ER) is true, veridical and non-veridical empirical experiences are of radically different kinds. I have not denied that it is possible to have non-veridical experiences that seem to be of external objects; I have not denied that there are dreams, hallucinations, and so forth. But while the epistemically available aspects of veridical experience are fully fused to the objects they are experiences of, seeming to experience a pig is not fully fused to a pig. Seeming to experience a pig does not fit into schema (FA2) under epistemically available aspects, even supposing that it is a genuine relation. (However, if seeming to experience a pig is a relation, it might satisy (FA1); it might be impossible, e.g., to seem to see a pig if the environment contained no pig-like features whatever.) The epistemically available aspects of non-veridical experiences are not fully fused to anything, though they may be partially fused. Putting it very roughly, dreams and hallucinations are in some sense internal to the cognizer while veridical empirical experiences are not. So, to repeat, veridical and non-veridical experiences are radically different.


Now such a view runs afoul of an ancient argument, an argument that will doubtless already have occurred to the reader. That argument runs as follows:







(1) The experience of hallucinating a pig may be mistaken for the experience of seeing a pig.


(2) Thus, what is going on in the experiencer during these two experiences are closely similar, or even "qualitatively identical."


(3) Since in both cases one ends up taking oneself to be seeing a pig, in particular the epistemically available aspects of the experience of hallucinating a pig are the same (or very similar) to the epistemically available aspects of seeing a pig.


Therefore, (4) Internalism with regard to veridical experience is true.



Now such an argument has an intuitive appeal, as can be seen from its persistence in the history of philosophy. Indeed, some such argument underlies Cartesian scepticism, and thus, as I have argued, it must to a large extent inform the history of the epistemology of justification. But I think it is obviously a very weak argument. It appears to rely on some form of abductive inference, some principle to the effect that if two things have certain similarities, they have other similarities, or even a principle to the effect that if two things have certain similarities, they have some identical component. Obviously, no inference that appeals to such a principle could establish its conclusion with certainty. An android may look for all the world like a human being, yet not be, all things considered, very much like a human being at all; it may have no component which is closely similar to any component of a human being.


Premise (2) is clearly problematic. It seems to appeal to a principle to the effect that we cannot mistake things for one another unless they are closely similar or of the same kind. That principle is blatantly false. Furthermore, the proponent of (ER) can happily admit that hallucinating a pig and seeing a pig are similar in certain respects; similar firings of neurons might take place in both cases for example. Thus it must be established that certain of the similarities are epistemically available. Premise (3) purports to do this, but again, it appears to rely on some extremely tenuous form of abduction.


Even were we to admit that the argument is valid, however, it begs the question against the proponent of (ER). The argument assumes that internalism is true with regard to veridical experience in order to show that it is. To repeat, it is argued that such experiences as seeming to see a pig may be "qualitatively identical" or "indiscernible" from experiences of seeing a pig. That is, it is possible to be appeared to pigly without a pig being present. It is then argued that the similarity between seeming to see a pig and seeing a pig suggests that there is some component in common to the experiences, that when one sees a pig, one also seems to see one.


But we can admit all the facts mustered for this conclusion and deny the conclusion. Now one rather weak argument to the effect that "qualitative indiscernibility" provides grounds for the conclusion derives from Ockham's Razor. That is, since we can give a more or less univocal account of the two sorts of experience on the internalist view, it might be held that this fact counts in favor of the view. About this I would merely like to say that although we should not needlessly multiply entities, we ought to acknowledge all the entities there are that need explanation, and that prima facie we seem to recognize a deep and important distinction between veridical experience and hallucination.


The argument could perhaps be re-cast to avoid some of these problems. But, I claim, any way of re-casting the argument will be question-begging against (ER), because any such re-casting of the argument is going to have to appeal at some stage to "qualitative identity" or to an internalistic sense of indiscernibility. These notions smuggle seemings or appearances into the argument in a way that assumes that internalism is true. For qualitative identity and indiscernibility (as it is used in the argument) are notions of apparent similiarity. If two things are qualitatively indentical, this is just to say that they seem identical; if two things are indiscernible in the sense required in the argument, this is just to say that they appear the same. Certainly, it cannot be claimed that, all in all, the experience of hallucinating a pig is indiscernible from the experience of seeing a pig. This is because seeing a pig requires the presence of a pig, and a situation in which there is a pig present is not indiscernible from a situation in which there is not a pig present. Of course, I may be deluded into thinking there is a pig present when there is not. But if we take this delusion as a mark of indiscernibility between the two experiences, we a pre-supposing the legitimacy of isolating a purely internal, epistemically available aspect of veridical as well as hallucinatory experience: just what the argument purports in this context to prove. The argument presupposes that (ER) is false, that the epistemically available aspects of veridical experience are not fully fused to the object of that experience. If (ER) is true, then we cannot isolate an internal point of view in the first place.


The illegitimate move is from the possibility of seeming to see a pig when no pig is present to the claim that seeming to see a pig is an epistemically available aspect of actually seeing a pig, and further, to the claim that a pig causes me to seem to see a pig, and to the claim that I infer that I see a pig from the "fact" that I seem to see a pig. The experience of seeing a pig is then analyzed as (a) seeming to see a pig, and (b) a pig being where I seem to see it (along with some causal condition). But why should it be analyzed this way? Why not allow dreams, hallucinations, (as well as pains) and so forth, to be isolated as fundamentally different sorts of experiences than experiences of seeing a pig? Seeming to see a pig, or its seeming to me that I see a pig is not, if (ER) is true, an aspect of seeing a pig. Or, again, if veridical experience of a pig satisfies (FA2) under its epistemically available aspects, it is not the case that the pig causes me to display certain instrinsic properties, from which I then infer the presence of a pig. In the case where there is a pig under my nose I do not infer from the "fact" that I seem to see a pig that there's a pig. I simply see a pig. It is the pig itself under some of its aspects, which is epistemically available; nothing internal to me is epistemically available.


Of course, I may at times erroneously infer from seeming to see a pig that there's a pig, but I need not, and we need not take such a circumstance as the starting point of an account of empirical experience. Or at the very least, much more of an argument is required than has been provided by proponents of the internalism that we ought to make this our starting point. Again, we can admit all the data involved in hallucinations, dreams, etc., and we can admit that hallucinations are sometimes mistaken for realities and even that certain aspects of hallucinations may be self-presenting. But what we need to know is, why should these facts be taken as a starting-point for an account of what happens when I am not hallucinating? Why should we start from a position internal to the cognizer? Now surely the answer to this is going to be something like the following: if the cognizer is going to be justified in her beliefs about the external world, it is going to have to be in virtue of facts to which the cognizer has access (this is the classic objection to epistemological externalism). And since from the point of view of the cognizer the experience of hallucinating a pig may be indistinguishable from seeing a pig, we cannot start with a description of the situation from an external point of view if we are to get an account of the cognizer's own justification going. I am going to consider such claims about justification in detail later, and it should be noted that these claims would beg the question against the main thesis of this book: that knowledge does not require justification. But, for the moment, even if knowledge does require justification, we can point out that this account of justification pre-supposes exactly what I am claiming it needs to show, that is, that in a case of seeing something under our noses, there is a purely internal, epistemically available aspect of the experience that can be isolated; it pre-supposes that we can usefully describe the experience at all from a "purely internal" point of view; it pre-supposes that we can bifurcate veridical experience into something like an hallucination plus a state of affairs in the world. It is incumbent on the internalist to show that the appeal to seemings, appearances, and so forth is even possible, that is, to show that, e.g., I am appeared to in any way at all when I see a pig, before he argues empirical beliefs must be justified "internally." There are, after all, competing pictures of how such claims could be justified.



IV. Against Internalism


If (ER) is true, the epistemological project of studying what seems to the individual to be happening within him when he in fact experiences external objects, is radically impossible. There are, in typical cases of empirical experience, no "phenomena" to isolate and study. My experiences do not have any epistemicaly available aspects which are properties of me considered in isolation from my environment; they are, rather, properties of a situation in which I am a component. To consider how it is "with me," inside me, as I experience a pig, is, as an epistemological strategy, false not only to what is actually happening (I am in interchange with my environment, am situated with respect to it, am in spatial juxtaposition with a pig, literally in contact with it by means of my sense organs), but also to what I take to be happening (I take it that I am spatially juxtaposed with a pig, and so forth). Of course, I can have "phenomenological" experiences; I can have pains, delusions, hallucinations, deceptive dreams. Some aspects of these may well be "self-presenting." But the extraordinarily odd claim is that I should try to describe cases of actually seeing a pig by appeal to cases of hallucinating a pig (and so forth), rather than vice versa.


That is why a internalist of the kind we are considering cannot be engaged in a reconstruction of what we actually do. We do not, as a matter of fact, start off with "phenomena" and infer features of reality. Rather, we start with reality, situations in which we find ourselves, and, if we are philosophers of a certain kind, infer (erroneously for the most part), features of phenomena. And if the epistemically available aspects of an veridical experience are aspects under which the experiential relation is fully fused to the object that is experienced, we could not possibly start with "phenomena," because there are no phenomena, if the experience is veridical.


This view about experience is not wholly new. I take it to make somewhat more precise the view of Dewey, for example. As Dewey saw, experience is an interchange between organism and environment. He writes:



`Experience' denotes the planted field, the sowed seeds, the reaped harvests, the changes of night and day, spring and autumn, wet and dry, heat and cold, that are observed, feared, longed for; it also denotes the one who plants and reaps, who works and rejoices, hopes, fears, plans, invokes magic or chemistry to aid him, who is downcast or triumphant. It is `double barrelled' in that it recognizes in its primary integrity no division between act and material, subject and object, but contains them both.(6)



This passage, I think, contains the seeds of an alternative to the epistemology of justification, precisely because it countenances what I have been calling fusion. The epistemology of justification, I hope to suggest, requires a bifurcation of the epistemic agent from the world. But if Dewey is right, this bifurcation is invidious; we, as epistemic agents, are already part of the world. As epistemic agents and as human beings, we are part of the natural system; we are fused with it. We participate epistemically in the world in just the same ways that we participate in other ways with the world, as ourselves aspects and segments of that world, as participants rather than observers.


In a certain sense, and as I think Dewey would argue, we can get misled by starting with visual experience, for here it is at least not baldly absurd to separate inner and outer components; it is at least understandable how someone might think that such a separation makes sense in typical cases of seeing an object. But consider other kinds of experience: for example the experience of breathing or of eating. Here the bifurcation of experience into inner and outer components is obviously out of place. Breathing and eating involve actually incorporating bits of the world into one's body; you simply cannot describe a process of eating by restricting your attention to what is internal to the person who is eating. Eating involves appropriating something from the environment and incorporating it into one's body; it is an interchange between organism and environment. To a certain extent, I believe that it is a compelling picture (at least as compelling as the bifurcated picture) of vision, for example, or hearing or touch, to describe such experiences by analogy to breathing or eating. Visual experience is a situation of an organism in an environment in which elements of the environment are appropriated by the organism. There is no accurate description of an actual visual experience which is a description of states of the organism in isolation from that environment.


Some people, we might conclude, are appeared to pigly, others have pigs thrust upon them.



V. Two Senses of `See'


Ultimately, a decisive demonstration of (ER) awaits a fully developed science of human experience. Whether (ER) is true seems to me to be ultimately an empirical question. Nevertheless, I think support for the view arises from our everyday practices and talk with regard to experience. In evaluating the extent of this support, I will return to the rather hoary debate between Austin and Ayer on perceptual knowledge. This debate, though in many ways superseded, still strikes me as broaching in an exemplary way the basic arguments and the basic range of possible positions.


At any rate, in Sense and Sensibilia Austin describes a distinction introduced by Ayer between two senses of `see': in the first "it is necessary that what is seen should really exist, but not necessary that it should have the properties it appears to have."(7) In the other sense "it is not possible that anything should seem to have qualities it does not really have, but also not necessary that what is seen should really exist" (p. 94). As Austin points out, there just is no such sense of `see' as the second in ordinary language. I cannot truly say that I see a pig when I'm standing in an empty field, and then appeal to this different sense of the term. We must remark that Ayer's way of formulating the purported distinction stretches (roughly, at any rate) at least to Descartes, where it is introduced more generally with regard to `perceive.' But it is no more true that there is in ordinary language such a sense of `perceive' than that there is such a sense of `see.' And we have, of course, made the very same point about `experience' in contexts of the form `S has an experience of x': it is impossible to have an experience of what is not there.


Now I have fewer reservations than Austin about introducing technical senses of these terms; I would allow Ayer to introduce whatever sense of `see' and so forth he deems fit and finds useful. Of course, the question arises whether anything actually satisfies the stipulated definition. If (ER) is true, then nothing does satisfy the stipulated conditions in the case where we are actually experiencing some external object. Of course, I have done no more thus far than suggest that some of the obvious attacks on (ER) can be answered. For the moment, however, I will not assume that (ER) is true. The problem I want to press here is this: the internalist with regard to veridical experience introduces a technical notion of `seeing,' `perceiving,' `experiencing,' and then, seduced, as it were, by his own stipulations into thinking that seeing in his technical sense has something to do with seeing in the ordinary sense, he re-introduces his stipulated meanings back into ordinary contexts. Then he concludes that two things are going on when I see in the ordinary sense a pig. That is, I see in the ordinary sense that entails that there is a pig and in the technical sense which does not entail that there is a pig, but entails that I see something with the qualities it appears to have. This is similar for our purposes to Chisholm's claim that visual appearances are self-presenting. And here as in Chisholm, the question is supposed to be how I move from appearances to reality, from seeing in the stipulated sense to seeing in the ordinary sense. Without considering the question of whether these are true claims about vision, or experience, we can at least see that such a move is not forced on us even if we do allow that some "visual experiences" (e.g. hallucinations) are self-presenting, or that some events satisfy the stipulated conditions.


Here is a place where I think Ayer makes the sort of mistake just described, where he re-introduces a stipulated sense of `experience' back into an ordinary context without providing a separate argument that this move is legitimate:



[T]hose who say that even the most straightforward judgement of perception like `This is a table' embodies an inference are trying to make exactly the same logical point as those who say that all such judgements are uncertain. In both instances the contention is that the judgement goes beyond the data on which it is grounded, that it claims more than is contained in the experience which gives rise to it, that it makes assumptions which may be false, consistently with this experience.(8)



Here we move, inexplicably, from the the stipulation of a sense of `experience' on which I can have an experience of seeing a table without a table actually being present, to the conclusion that experience in the stipulated sense is a component of an ordinary experience of a table. And note that Chisholm makes an analogous move when he equates "being appeared to" with "sensing." `Sensing' is in this respect just like `experiencing.' `S senses x' introduces an extensional context, and so forth. Chisholm must be relying on a sense of `sense' analogous to Ayer's sense of `see,' and we can see that, in fact, the cases are analogous, because "sensing" in Chisholm's sense is "self-presenting." We assume now that the `experience of a table,' or the property of sensing a table, can be neatly lopped off from the table. But this is true only in the stipulated sense, and no reason has been given to think that the stipulated sense is in play at all in the ordinary claim that this is a table. To put it in Chisholm's terms, no reason has been given to think that there is a component of the experience that is self-presenting. In the most general terms: no reason has been given to think that there is an epistemically available component of the experience that is intrinsic or internal to the experiencer and which is necesary for the experience in question. (As an aside, it will be obvious from the previous discussion that I think that what Ayer says is false in this sense: the table is contained in an experience of it; the experience of a table is a state of affairs in which the organism is fused with a table.) If it is illegitimate to assume that the stipulated sense of "experience" and similar verbs is in play in the normal case, it will a fortiori be illegitimate to assume that beliefs about the experienced item can be justified by their relations to aspects of the event that supposedly satisfies this stipulated sense. So far, in fact, we have seen no reason to think that such beliefs can be justified at all.


Under one construal, the question comes down to what `data' might mean here. Ayer obviously means sense data. But I do not see why the table is not itself among the data in an experience of a table. This is not, again, to say that I am infallible with regard to the sensory detection of tables. Nevertheless, it is false to hold that the fact that I might be wrong shows that I am engaged in an inference that goes beyond the data, as Ayer and Chisholm suggest. The way I have gone wrong is not necessarily that I have misinterpreted the data in a certain sense as corresponding to the way things are, but simply that I have gone wrong about what data are given in the experience, so that I think, wrongly, that a table is such a datum. I think it is the fact that they miss this point that misleads internalists here. That is, I could be mistaken about the presence of the table even if I am engaged in no inference at all regarding (and no activity at all, for that matter, in which I am engaged with) internal, epistemically available components of my experience. I may simply go wrong about what I am seeing, without this fact showing that I am engaged in a mistaken inference. There are, after all, various mistakes I can make that are not inferential mistakes. (For example, I run a stop sign, not because I have mistakenly inferred that there is no stop sign, but simply because I fail to see that the stop sign that is there.)



VI. Strict Seeing


This brings us to another of Ayer's points about the supposed inferences involved in "judgements of perception":



[I]n identifying something I see as an apple, I assume not only that it is tangible but that it has a certain characteristic texture. I make assumptions about the way it smells and tastes and about the material out of which it is made, for example, that it is a fleshy fruit and not an object made of wax. I may also be assuming something about its origin, for example that it was grown on a tree, and about its causal properties. I assume further that it has other faces than the one which is turned towards me, that it has an inside, that it is not hollow. But can it seriously be maintained that I see all this? Of course it is perfectly correct for me to say that I see the apple: but this just proves that in making a statement of this kind I commit myself to the existence of much more than I strictly do see. (p. 292)



Here again a technical sense of `sees' has been imported unjustifiably into ordinary discourse. For the question of "what I see" just would not usually be answered in the way Ayer answers it. I do not go around saying (or thinking) "Now I see the surface of one side of an apple." Here we are dealing with yet a third sense of `sees,' which Ayer terms "strict" seeing. But "strict" seeing is similar to the previous stipulated sense in that it takes a view internal to the experiencer to be primary. If one "strictly sees" x, it does not seem to follow that x has the properties it seems to have. So strict seeing is not self-presenting. Nevertheless, strict seeing is a philosophical construct with no analogue in ordinary usage. It is a philosophical fragmentation that is not endorsed by our usual ways of talking about experience. What one strictly sees in the ordinary sense, I have been arguing, is an apple with such-and-such a texture, such-and-such a taste, etc. I might infer that it has such properties, or I might not, but it does have them, in the described case. Here again, experience is lopped off from the world in an unwarranted way: we cannot accurately describe this experience except as an experience of the apple, the apple that also has such-and-such properties. To say that I "strictly see" little bits of the apple is to suppose that a stipulated sense is in play in the ordinary case. It is to assume, in fact, that a version of internalism with regard to veridical experience is true. And to establish this a positive argument is required.


Here, I think Ayer confuses inference with what Wittgenstein would call ways of acting. If I see an apple, or think I see an apple, and am hungry and like apples, I may pick the thing up and bite it, or try to. If it is wax, I will be surprised and disappointed. But this does not show that I am engaged in drawing inferences. It only shows that I act in certain ways in the presence of what I take to be apples. The claim `this is an apple' may be "proleptic" in Ayer's sense; I may think: eat it; I may think: toss it at that chap playing Hamlet. But I need not have any expectations at all. However, it must be admitted that we assume (and believe enduringly) that something we take to be an apple grew on a tree, and so forth. This, however, obviously does not show that we infer from anything at all that it is an apple. That it is an apple entails that it grew on a tree, and I believe this enduringly. But this hardly counts as evidence that it is an apple. Note that `I am being appeared to appley' also has entailments which "exceed the present data": that I exist, that I have an idea of what an apple looks like, that the universe is such that experiences of this kind are possible, and so forth. But this, Ayer has to hold, does not show that I have inferred how I am being appeared to from anything at all.



VII. Square Tables


Another very old argument (stretching back at least to Berkeley) is frequently brought to bear in the project of showing that we infer the properties of external things from appearances (and an analogous argument might be formulated to show that appearances make the properties of things evident to us in Chisholm's sense). The argument goes roughly like this: a square table, for example, does not in fact appear to be square from any angle except from at a certain orientation directly above or below. From different angles, the table "projects" the appearance of various distorted parallelograms. We infer, it is held, that the table is square as a simplifying hypothesis, as a way of bringing its many geometrical appearances into some sort of unity. But my reply here will be familiar. Nothing constrains us to admit that the table appears to be different shapes from different angles. Now of course, and again, we can make mistakes about the shape of things. But such mistakes do not have to be construed as arising from appearances of any sort; in fact, they are most naturally construed as just what they seem (!) to be: mistakes about the shape of some external thing, mistakes about what data are given in experience.


We should pause for a moment to appreciate how deeply counter-intuitive the account in terms of a variety of geometrical appearances is here. If we are in the presence of a square table, staring at it in good conditions from an oblique angle, and we are asked what shape the table looks to have, we answer straightaway: "square." It is anything but evident, that is, that the table does project distorted parallelograms, and so forth. If we were to respond that the table looks to be shaped as a distorted parallelogram, eyebrows would certainly be raised. We can see that it is square.


Again, ordinary language carries with it not a demonstration of the falsity of the position, but a burden of proof. The advocate of appearances is appealing to "facts" which are anything but evident, and needs an argument to show that there are such facts. Perhaps the answer here will be: the hypothesis is a simplifying one; it has explanatory power. But what does it explain? How we get from appearances to reality? That is a question that is manufactured by the view in question; the view scores no points for solving a problem that does not arise except on the condition that the view is true. And as an explanation, it does not have the virtue of simplicity; indeed, it is rather baroque. For the view with which it competes is that under decent conditions the table looks square from any angle (though again, mistakes are not ruled out).


But it will be felt that I am not doing justice to the initial intuitive plausibility of the internalism with regard to veridical experience in cases such as that of the table as seen from oblique angles. Well, I admit, I am immune to this particular intuition. But we might tell a story of how this view gets going by considering the fact that in classical empiricism, visual experience is often thought of by analogy to depiction. It is not hard to see how, if this analogy is carried far enough, it will suggest that visually discernible items "project" different shapes onto our visual apparatus from different angles. The analogy here is, roughly, to perspective rendering. If you want to draw a table from a certain angle, you had better foreshorten it in certain systematic ways; what will appear on the paper will be a distorted parallelogram, and this parallelogram will then be interpreted as, will be inferred to be, a depiction of a square table. But one way to understand the present objection to internalism with regard to experience is that this analogy of vision to depiction is illegitimate.







VIII. The Practice of Justification


I have argued that we need not justify, in a general way, our beliefs about the external world, or, better, offer a general style of justification, applicable to all such beliefs. I would now like to supplement this claim by arguing that in certain specific cases we need not or, perhaps, indeed cannot offer any such justification, if justification is held to be internal.


It is common for advocates of internalist conceptions of justification to say that what they are doing is not merely reporting the actual psychological situation we are in when we are asked how we know. They take themselves, rather, to be offering a "rational reconstruction" of the situation; they are trying to say how people might on reflection justify their beliefs. Thus, "It seems to me now that I am seeing a flesh-colored patch," or "I am appeared to flesh-colored patchedly" or merely "thusly" are produced as examples of how we typically would, or at any rate (because otherwise the claim is baldly ludicrous) how we might justify the claim that "here is a hand." That is, we justify claims about the external world by appeal to internal or phenomenal states of affairs. I will argue that such justifications not only do not bear a "reconstructive" relation to our practice, but that they are fundamentally divorced from practice, that they bear no relation to practice whatever.


It seems to me that when, to repeat an observation from the last chapter, as I raise my hand and gaze at it, I say "here is a hand," I am absolutely at the level of foundations. "Here is a hand" in the circumstances Moore produces it as as certain as any claim could possibly be. The idea that there is, or that there needs to be, evidence for such a claim seems to me bizarre. If it makes sense to say "it appears to me now that I am seeing a hand," is that claim more probably true than the claim that "here is a hand?" On one picture of justification, we ought to adduce more certain claims in support of less certain ones. But if I were to acknowledge the propriety of claims about appearances in this context, I think I would offhand assign either of the above claims a probability very close to 1. Of course, the view is going to be that claims about appearances are or may be self-presenting, self-justified, incorrigible, and so forth, presumably a status that entails that they are more probable or more certain than empirical claims about middle-sized physical objects. But it is hard to see how moving to "appearances" etc. could make us any more certain than we are already, without any such appeal, that here is a hand. But then again, the internalist claim may be that we are already, as it were implicitly, appealing to appearances when we make the initial claim that here is hand, that it is appearance that makes it evident to us that here is a hand. That may be the sense in which the internalist is claiming that his account is a reconstruction of our practice. Indeed, I do not see how else to interpret the internalist claim, because I do not see how I could be any more certain of anything that I believe than I am about "here is a hand" in the described circumstances. So it is this claim that I want to focus on, the claim that doxastic agents are already in the business of inferring reality from appearances, or at any rate moving in one way or another from appearances to reality. And I will argue that there is at least no compelling reason to accept the view, and that an alternative conception may possess much more initial plausibility.


We shall come back to appearances and so forth later. For now, let us return to the initial claim that I can and should give evidence for simple empirical claims. For the internalist about justification obviously cannot rest content with the mere assertion that I move from appearances to reality; he must also assert that this move confers justification on certain of my empirical claims. Justification is, at least in part, and on an internalist conception, a matter of being in a position to give reasons for a belief. Thus, the internalist is obliged to hold that, if I am reflective, I can give reasons to think that an empirical claim that I know to be true is true by an appeal to appearances (seemings, sense data, phenomena). It is this claim that will be the subject of discussion: the claim not merely that I do move in some way from appearance to reality, but that, in order to be said to know some simple empirical proposition, I can and must be able to give reasons, and specifically reasons that are framed in terms of appearances and so forth.







IX. Austin's Argument


Austin gives a famous battery of arguments in Sense and Sensibilia against the claim that I need to, or indeed can, give evidence or reasons for certain claims which I know to be true. He considers the example, already introduced above, of a case wherein I claim "that is a pig" when a pig is in full view before me:



The situation in which I could properly be said to have evidence for the statement that some animal is a pig is that, for example, in which the beast itself is not actually on view, but I can see plenty of pig-like marks on the ground outside its retreat. . . . But if the animal then emerges and stands there plainly in view, there is no longer any question of collecting evidence; its coming into view doesn't provide me with more evidence that it's a pig, I can now just see that it is; the question is settled. (115)



I do not give evidence, or have evidence, or have to provide evidence for what, in Austin's phrase "is going on under my nose." That I am certain about what is going on under my nose, that I know it, is the sort of situation that gives the terms `certainty' and `knowledge' their meaning. People do not go about satisfying themselves by examining the evidence for "here is a hand," because nothing they produced could make them more certain than they already are. Again, this does not provide an attack on the claim that empirical propositions are inferred from propositions about appearances, or that it is appearance that makes reality evident to us. But it does attack the claim that simple empirical beliefs held under good conditions need to be justified by an appeal to appearances, or indeed in any way at all compatible with internalism, in order to count as knowledge. Now it might be asked why this should concern the internalist. Well, it should. For either the question is about how we know such things as that here is a hand or here is a pig, or the question is really of no interest. The internalist view, extended to include a view about justification, might be construed as a program to revise our practice, or to institute a new practice; only then it is incumbent on its proponent to say exactly what is wrong with our current practices. But as it stands, we do no not, as I have been and will be arguing, justify the claim that here is a pig by appeal to appearances in the described case. That is, the internalist account of justification in cases like this bears no relation to our actual practices: we do not, usually, provide justifications in such cases, and certainly we do not, as a rule, provide justifications by appeal to appearances. The internalist is obliged to show either that we do justify such claims in such ways, or renounce the claim to be giving a reconstruction of practice. If he does renounce the claim to be giving a reconstruction of practice, then he is obliged to say what is wrong with our current practices of justification: why we should, and how we could, institute a new practice. In addition, the internalist program will now be profoundly revisionary about the current scope of human knowledge. That is, when people simply wave off the demand for an internalist justification for "here is a hand" and the like, when they are both unable and unwilling to produce a justification in the envisaged style, the internalist is going to have to hold that they do not know the simple empirical propositions they believe. I suppose I would regard it as an adequate answer to any account of knowledge that some normal human subjects do not know that "here is a hand" when they are staring at their own hand in broad daylight. (The reasons why I regard this as an adequate answer are set out in the chapter two.)


Now of course, it is likely that, given the fact that there are possible defeaters for the claim that here is a pig, the foundationalist will argue that, if indeed we cannot justify the claim, then we don't know it to be true. But to begin to see what is wrong with this view, consider how we would actually respond to the demand to justify the claim that "here is a pig" when there is a pig under our noses. To begin with, the demand is eccentric; it sounds like an aspersion on our sanity; we would probably respond with a raised eyebrow or an expression of exasperation or verbal abuse. But of course, these aren't justifications (at least not in the epistemological sense of `justification'). If for one reason or another we decided that a serious response was in order, we might simply gesture at the pig, or say "look!" But this can hardly be construed as an appeal to appearances, at least not without some further argument. Rather, it seems that we are, roughly re-iterating that this is a pig, which is to reject rather than satisfy the demand for a justification. In our practices, we do reject the demand for a justification in such cases, or to be more accurate, in such cases the demand for a justification simply does not arise.


Austin considers another, similar case:



If, for instance, someone remarks in casual conversation, `As a matter of fact I live in Oxford,' the other party to the conversation may, if he finds it worth doing, verify the assertion; but the speaker, of course, has no need to do this - he knows it to be true (or, if he is lying, false). Strictly speaking, indeed, it is not that he has no need to verify his statement; the case is rather that, since he already knows it to be true, nothing whatever could count as his verifying it. Nor need it be true that he is in this position in virtue of having verified his assertion at some previous stage, for of how many people really, who know quite well where they live, could it be said that they have at any time verified that they live there? When would they be supposed to have done this? In what way? And why? (119)



One need not agree with Austin's "ordinary language" view about such terms as `verification' and `evidence' to take his point. Austin's point is usually said to be that it only makes sense to talk about `verification' and `evidence' where there is an open question, a genuine doubt. In Ayer's reply to Sense and Sensibilia, I think he gets the better of this debate when he says "If I think I know that p, I am underplaying my hand, and so misleading my audience, if I say no more than that I have good evidence for p. It would, however, be rash to lay any weight upon this in the present context, since my knowing that p is certainly not inconsistent with my having good evidence for it" (289). This is certainly right. But even if we allow that I could have evidence about where I live and could verify where I live, the point is that I do not know where I live in virtue of any particular body of evidence, and certainly not in virtue of having been appeared to in any particular way. In some ways, this is a better and more typical case of empirical knowledge than "here is a hand (pig)." In the case of "here is a hand" one can understand the temptation to try to justify the claim in terms of appearances. "Here is a hand" is neatly indexed to a time and place; it reflects a momentary state of affairs. If there were any appearances bouncing around in this case, there would be a neat little appearance bundled up with the claim. But in virtue of what appearances might I know where I live? In virtue of the whole series of appearances I had as I approached the place, negotiated with the landlord, moved in, and then failed to move out? If there were relevant appearances here, they would include more or less the entirety of the ways I have been appeared to over the last x years. I might say that these appearances justify my claim. But certainly if you demand that I produce my justification, if you ask me how I know I live at . . . , I simply cannot respond. Or if I respond by saying it has to do with all the ways I've been appeared to over the last x years, how is this supposed to be regarded as a justification of this particular claim as opposed to an indefinite number of other claims? And doesn't this whole exercise begin to look bizarre, completely unrelated to how anyone actually does justify anything?


And of course, if I have to justify the claim that I live at . . . in virtue of all the ways I have been appeared to over the last x years, then I am no longer appealing to anything that can plausibly be held to be self-justified or self-presenting. Obviously I am now subject to mistakes about how I was appeared to years ago, and I am perfectly aware of this fact. How do I know that I have not in fact been appeared to in ways that contradict the claim that I live at . . .? After all, I hardly recall all the ways I have been appeared to in the course of a period of years, or anything but a tiny portion of them. Am I less sure that I live at . . . than that I have not been appeared to in a relevant way some time in the past? (And consider some examples from Wittgenstein, e.g., "I have never been to the moon.")



X. Wittgenstein and Epistemological Bedrock


To return to the claim that here is a pig: we need some care here. Even if an appeal to appearances is precluded as bearing no relation to practice, I could produce propositions from which I could infer that here is a pig, and it is worth taking a moment to explore the question of whether these could be construed as justifications for the claim. (Here we momentarily leave behind the classical internalist program.) For example, I could infer that here is a pig from such claims as "there is a pig in that pen," "I am now looking at a pig," "the pig is oinking," etc. I don't have to make "here is a pig" my epistemological bedrock. The point is that, when I decide to stop justifying in this way, rather than by a retreat to appearances, I am going to end up with a proposition with the same general status: a contingent, corrigible, empirical claim. As Wittgenstein puts it in On Certainty, I must "begin to trust somewhere."(9) That is, I am going to have to stop the process at what I have been calling faith. (That I do not know the proposition I believe on faith only follows if justification is logically necessary for knowledge.) If I am willing to start justifying at all I am going to have to stop somewhere. And if I stop without appealing to apearances I will have an eminently corrigible proposition which I treat as `logical' in Wittgenstein's sense, as something I exempt from doubt, as a rule of the game. (Nevertheless, though I exempt this claim from doubt, it is, as I pointed out in the last chapter, a perfectly normal empirical claim: a claim about the world. That is why I myself would prefer not to call it "logical." Empirical evidence, for example, could certainly count against such a claim, in a way that empirical evidence could not, e.g., count against a rule of a game.) The proposition `Here is a pig' can serve as bedrock because, again quoting Wittgenstein (who in turn derives the point from Moore), it "is as sure a thing for me as any grounds I could give for it" (#111). That is, I don't gain any purchase by trying to justify "here is a pig" with "the pig is oinking"; I don't improve my epistemic position. I might as well stop here as anywhere.


Now Wittgenstein stops short of Moore's conclusion that I know that here is a hand (Cf. #151); this seems bound up with his claim about the status of such propositions as "rules of the game." But I don't see why we should deny that these are cases of knowledge, except that we are already in the thrall of the epistemology of justification. Indeed, I know that here is a hand as well as I know anything. Wittgenstein asks "Can I believe for one moment that I have ever been to the stratosphere? No. So do I know the contrary, like Moore?" (#218) And again, "I should like to say: Moore does not know what he asserts he knows, but it stands fast for him, and also for me; regarding it as absolutely solid is part of our method of doubt and enquiry" (#151). These remarks seem to me to make too much of a concession to the epistemology of justification. As Wittgenstein might put it, "know" has migrated out of the language games in which it has a use. Or rather, the situation is this: Wittgenstein has allowed `know' to be appropriated by the epistemology of justification, to the point where we don't `know' what is not absolutely metaphysically certain, the point at which to know entails that one can show decisively that one could not be wrong: "`I know' relates to the possibility of demonstrating the truth" (#243). It seems to me, though, that, if we choose to, we can hold on to the word `know'; it need not become merely a philosopher's term of art. There are ongoing practices in which `know' is a perfectly functional piece; I know perfectly well (and not in any qualified sense) that I have never been to Mars.


Again, one test of whether I know such a proposition and whether it is part of my epistemological bedrock is not that I can produce a justification (say, recite my itinerary for the last thirty-four years). The test is rather that I reject the demand for justification out of hand, that the demand is absurd. Only someone truly in the grips to a theory would take such a rejection to show that I must be justified after all. Of course, in some cases where I do not produce a justification, I could, if pressed. And in some such cases, the justification of my belief might also describe its source. For example, I might know that the car is out of gas because the gauge points to `E,' and I might look at you funny if you demanded a justification. But the point here is that the demand is appropriate epistemically; the claim is of a kind for which people normally do have and are willing to produce evidence. That is not the case with the claims I am now considering. I am not saying, when you ask me how I know that I have never been to Mars, that I am so obviously justified that I don't feel like producing my justification; I am saying that justification is not necessary, that the demand is inappropriate. I am rejecting meeting the demand as a reqirement to show that I do know. The demand should not arise, unless a recherche sense of knowledge is in play.



XI. An Objection


Here is one argument that is often put forward by internalists with regard to experience (extended into an account of justification) in order to show that their view is a reconstruction of practice. We might justify the claim that "There's a pig," we might respond to the question "How do you know?" by saying such things as "It certainly seems to be a pig," "It appears to be a pig" and so forth. So although it is obvious that we would, in a variety of circumstances, simply reject the demand for a justification in such a case, it is not clear that we would always reject that demand, and, furthermore, it seems that we might, and occasionally do, respond to that demand by an appeal to appearances or seemings. This would hardly be enough to establish that we can always justify simple bits of empirical knowledge by appeal to appearances, or some internal, epistemically available component of experience. But it may suggest that the account is not so radically divorced from practice as I have been asserting it is.


However, I believe that this is not the most plausible way to construe these locutions ("It seems to be a pig," "It appears to be a pig"). To begin with, we can make the Austinian point that, for the most part, such locutions will be produced when serious grounds for doubt have been introduced, or when the conditions of experience are non-optimal. Otherwise, the demand for a justification will usually be rejected. But a second, though related point is this: such locutions have another obvious function in ordinary conversation besides a direct appeal to seemings or appearances. In ordinary conversation, "it seems that," and "it appears that" are often used simply as qualifications on a judgement, as expressing tentativeness about a belief. Such locutions often function in the same way that qualifications such as "maybe" and "perhaps" function. In other words, they may well operate over propositional beliefs rather than over experiences.


Consider the claims "I appear to live in Oxford," or "I seem to live in Oxford." Such locutions could not, without extreme strain, be interpreted as appealing to internal aspects of particular experiences. Rather, they express tentatively the belief that I do live in Oxford. We can see that this is the case because such locutions would be produced only in extremely unusual circumstances. I would say that I appear to live in Oxford only in a state in which I seem to myself to be fundamentally confused, in which I have been recently undergoing treatment for psychosis, and so forth. That is, I will apparently appeal to appearances only in those cases in which my judgement about where I live is a tentative one. The fact that such locutions are used to express tentative judgements is shown by the propriety of producing them in cases that run far afield of particular empirical experiences. For example, a scientist may say "Quarks appear to be quantum particles." This is best construed simply as expressing the tentative judgement that quarks are quantium particles. Certainly it is implausible to take it as an assertion about an internal aspect of my experience of quarks. Or again, think about: "The good appears to be identified with happiness."


I take it that the fact that "It seems to be a pig," and "It appears to be a pig" would typically be produced in cases where I suddenly find myself in doubt about my judgement that it is a pig, suggests that "seems" and "appears" are introduced as operators over propositional beliefs which express tentativity, rather than that they appeal to internal aspects of experience. The cases where it would simply not be plausible to claim that such locutions appeal to internal aspects of experiences (cases, e.g., where the entities referred to are abstract or unobservable) seem on the face of it entirely parallel to the cases wherein one would be tempted to interpret the locutions as retreating to the level of appearances. So I do not think that this argument even suggests that simple experiential claims can be justified by appeal to an inner component of experience.



XII. Foundationalism and Coherentism


I am, emphatically, not arguing that empirical claims made under good conditions - `here is a hand,' `there's a pig,' `I live in Nashville' - are self-justified in any sense, that, say, they satisfy the conditions on epistemological justification, but not in virtue of their relations to other propositions or states of affairs. I think that such claims simply do not meet the standards of epistemological justification in most of the ways that epistemological justification has been explicated. With claims such as "There's a pig" and "I live in Nashville," we have, as Wittgenstein puts it, arrived at the "rock bottom" of our convictions (#248). If we retreat to appearances and say that what gives evidence for "there's a pig" is the belief that a big pink blob is now apparently traversing my visual field, we are in the business of attributing bizarre beliefs and strategies for justification to people that they need not (and unless they are philosophers of a certain kind, do not) hold. And speaking only of philosophers, say that I do not believe that I can isolate facts about my visual field from facts about states of affairs in the world in cases of veridical experience. Say, in other words, that I am a sincere proponent of (ER). This obliges me to reject every ascription to me of beliefs about the sheer phenomenological contents of my mind in cases of veridical experience. I claim to have no such beliefs whatever. If we took such a declaration seriously, and if foundationalism of the sort that appeals to such contents to justify empirical beliefs were right, I would have no justified empirical beliefs. But surely I do know where I live, and so forth.


Philosophers aside, however, the fundamental point is that, in the circumstances in which Austin describes the production of "There's a pig" and "I live in Oxford," there is no way we can improve our epistemological situation with regard to them. They are as certain as any claims can be, including any claims that might be adduced in their support. That is why justification is out of the question.


I believe that roughly similar considerations will show that coherence theories of justification are inadequate. They are inadequate in the sense that they are profoundly revisionary of our practice, that they tend to show that people might not know many things which it is perfectly obvious that they do know, that in certain circumstances the view would not count as knowledge paradigmatic cases of knowledge, cases which give the term `knowledge' its sense in the language games of which it is a part. For a coherence theorist is likely to countenance as parts of the justification of the claim that here is a pig such claims as "The pig is oinking" and "The pig is in the pen." Various familiar objections to coherence theories have considerable bite here. For example, it seems on the face of it puzzling that claims which are no more certain than the belief to be justified could in any sense justify that belief. Again, we can pick our epistemological bedrock from among such claims, but it seems that our justification will run out. But the fundamental point I would make against coherence theory here, and indeed against any account of knowledge which requires that "Here is a pig" be justified when there is a pig under our nose, is simply the fact that, in normal conditions, the demand for a justification will be treated as completely unwarranted, as bizarre. In the next chapter, I will expand on this point. But for now, I want to repeat the conclusion of the last chapter: we are natural creatures with natural obligations. Epistemology ought to be construed as a reconstruction of the ways we conduct ourselves cognitively. If our practice entails waving off demands for a justification as bizarre in certain cases in which we do have knowledge, then the demand for a justification is out of place.



XIII. Conclusion


Of course, believing and knowing are in this respect exactly like seeing, perceiving, experiencing. That is, knowing and believing are not to be individuated as internal states of the epistemic agent. Whether or not (ER) is true, I take the present point to have been decisively established by Putnam-style "twin earth" cases. What one believes when one believes, say, that Lake Michigan is a body of water depends, in part, on the nature of the natural kind picked out by our term `water.' My double on twin earth, a planet on which `water' refers to XYZ rather than H2O, has a different belief than mine when the symbols `Lake Michigan is a body of water' run through his head. The truth of his belief entails that Lake Michigan consists of XYZ, while the truth of my belief entails that it consists of H2O.(10) Or consider knowledge. If knowing were an "internal" psychological state, there would be no distinction between knowing p and being extremely confident that p. But of course there is a difference: if I know that p, I can't be wrong. Obviously this has nothing to do with my infallibility. But we are liable to make the same kind of mistake with knowing that we make with seeing and so forth; we are liable to think that knowing contains an internal psychological component (belief) and an external, real-world component. That is, I know when I believe that p and it is true that p (with perhaps some added conditions). The problem is not with this general characterization; I certainly need to believe that p in order to know that p. The problem is, rather, with the characterization of belief as an internal state. Belief is something that, as it were, wraps its tentacles around the world, that reaches out and grasps the state of affairs with which it is concerned.


In the case of belief, I think, this quality of reaching out to the world (or the world reaching out to the belief) is parasitic on the referential function of the component of the internal representation by means of which we have the belief in question. To repeat an observation from the first chapter, we need to assure the reference for the believer of the elements of the representation under which the proposition in question is believed. And though I have argued elaborately that we do not need to postulate internal representations in order to understand seeing, perceiving, experiencing, and so forth, I do think that such a representation is required for enduring and occurrent belief. In any case, I accept, for reasons that may be apparent from the preceeding discussion, some version of a causal theory of reference. I will not pause to set such a theory out in detail, much less to mount a defense. But the causal theory might be put this way: to refer successfully to to an external-world entity E, I have got to be in the right sort of causal relation to E (the causal chain may well run through other speakers of the language from whom I acquire names and so forth). For just this reason, and for others I have been amassing, a theory of belief that tries to get going wholly through the syntax of my representations seems to me hopeless.


The fact that I can successfully refer to things outside myself using internal or external signs supervenes on my veridical empirical experiences. If such experiences are fused to their objects in the sense I have set out in this chapter, then the possibility of true belief depends on the fusion of the epistemic agent to the environment. If that is so, then I think in cases of garden-variety empirical beliefs, no justification in the classical sense is forthcoming. Nevertheless, we can know simple things about our environment. It follows that the epistemology of justification is radically inadequate to absolutely central cases of knowledge.


However, this conclusion, as we have seen, does fundamentally depend on an appeal to practice and intuition; it appeals to what we would say in certain cases, and so forth. And the epistemologist of justification, particularly if she takes the spectre of empirical scepticism seriously, may simply be willing to assert that practice deserves no weight here, or at any rate, that it deserves less weight than I have been giving it. She may simply assert that, even if it did turn out to be the case that most people did not know most of the simple empirical claims they claim to know, knowledge nevertheless requires justification. It is the purpose of the next chapter to mount a more systematic and "philosophical" attack on this last claim.


1. See, e.g., Alvin Goldman, Epistemology and Cognition (Boston: Harvard, 1986), p.154.


2. John Pollock, Contemporary Theories of Knowledge. (Totowa NJ: Rowman and Littlefield, 1986), p. 155.


3. Roderick Chisholm, "A Version of Foundationalism," reprinted in Human Knowledge, eds. Paul Moser and Arnold vander Nat (Oxford: Oxford University Press, 1987), p.298.


4. See Tyler Burge, "Individualism and the Mental," Midwest Studies in Philosophy, Volume IV: Studies in Metaphysics (Minneapolis: University of Minnesota Press, 1979), pp.73-121. Colin McGinn, Mental Content (Oxford: Basil Blackwell, 1979). Hilary Putnam, "The Meaning of `Meaning'," in his Mind, Language and Reality (Cambridge: Cambridge University Press, 1979).



5. McGinn, for example, writes: "Experience, we may say, represents the state of affairs so presented (or apparently presented) . . . The way in which experience represents the world constitutes its content, the way it makes things seem. The content of an experience determines what it is as of - how the world would actually be presented if the experience were veridical" (p.58). This is a version of the view I am rejecting.


6. John Dewey, Experience and Nature (New York: Dover, 1958 (1929)), p.8.


7. J.L. Austin, Sense and Sensibilia (Oxford: Oxford University Press, 1962), p.93.


8. A.J. Ayer, "Has Austin Refuted Sense Data?" Symposium on J.L. Austin, ed. K.T. Fann (London: Routledge and Kegan Paul, 1969), p.290.


9. Ludwig Wittgenstein, On Certainty (New York: Harper and Row, 1969), #150.


10. For elaborate arguments to this effect, see Colin McGinn, Mental Content (Oxford: Basil Blackwell, 1989).

















Chapter IV



Knowledge is Merely True Belief



The view I have been pushing so far is that there are cases of knowledge that are not cases of justified belief. I have argued elaborately that, in typical cases, beliefs such as "Here is a hand," and "I live in Oxford" need not be justified in order to count as knowledge. If radical externalism with regard to veridical experience is true, then an internalist conception of how such claims could be justified is false. (Later, I will have something to say explicitly about externalist conceptions of justification.) But the burden of the argument so far has been an appeal to practice; I have taken it as a test of a theory of knowledge that, if it turns out that on that theory, any typical person does not know where she lives, or that here's a hand, that demonstrates that the theory is false. That is, I have regarded typical claims to knowledge as being embedded in certain well-established practices, and I have regarded the test of accounts of knowledge as adequacy to such practices. I hope that I indicated, in the second chapter, why I do not think that this begs the question against scepticism (or perhaps it is more accurate to say that I indicated why it is alright to beg the question against scepticism). Furthermore, I have argued that knowledge rests on faith, that we cannot, in a global way, set about justifying our beliefs from the ground up. It is time now to turn to an explicit defense of the claim that knowledge is merely true belief, which, despite the claims advanced thus far, may still strike the reader as ludicrous. Indeed, though I have taken to task certain conceptions of justification that figure in the epistemological tradition, I have not yet given positive reasons to think that no justification condition (or indeed, no condition aside from truth and belief) is necessary for knowledge. I hope to repair this shortcoming in this chapter, to move on from criticism to a positive solution to the basic problem of epistemology.


I propose, then, the following analysis of `S knows that p':


(1) It is true that p.


(2) S believes that p.



Each of the two conditions is necessary; together, they are sufficient.


Now the reaction of a typical philosopher to this account is likely to fluctuate between amusement and irritation. Indeed, it is questionable whether, since the earliest discussions of the issue, any philosopher has explicitly taken the position I defend in here.(1) It is sometimes blithely claimed at the outset of a consideration of knowledge that our practices or our intuitions or our pre-analytic commitments show that we are already committed to a distinction between knowledge and true belief.(2) Now whatever may be the fact of the conceptual matter, I will argue that we are not pre-analytically committed to any such thing. Since no one has denied that some condition beyond truth and belief is required for knowledge, no one has offered an elaborate argument to the effect that some other condition is required. For the most part, we get a perfunctory appeal to our practices, along these lines: we do not count a lucky guess as knowledge, so something besides truth and belief is required. Thus, the project of the next several sections can be construed as an attempt to show that our practices are at best equivocal in this regard. In the final sections, I will move beyond an appeal to practice to mount what I regard as a decisive conceptual case for the view that propositional knowledge is mere true belief. But for the present, I will often appeal to "what we would say" in certain cases. This is not because I hold that ordinary language is invariably a good guide to meaning, or because I think there is no distinction between pragmatic contraints on utterance and semantic constraints on definition, but simply because the only arguments which appear to demand a serious response are framed in terms of linguistic practice.


Initially, I will not offer "proof" of the claim that knowledge is merely true belief, though I will eventually attempt to mount such a proof. What I shall try to do initially is suggest reasons why the burden of proof might be held in suspension. That is, I will begin by trying to answer the obvious objections to the view. Only after these are removed will I try to show that the view is correct.



I. Counter-Examples


In the first chapter, I treated belief as a propositional attitude. The reader will recall the definition of belief I gave there:



(D) S believes that p if and only if there is some item which functions for S as a representation of p and which S reflectively takes to express a true proposition.



To believe, very roughly, is to commit oneself to the truth of some proposition. On this and related pictures of belief, it is clear what strategy will be used to generate counter-examples to the claim that knowledge is merely true belief. Such examples arise in circumstances in which the connection between the belief and the truth of the believed proposition is not of the right kind.


Consider, then, the following counter-examples to the claim that knowledge is merely true belief:



(1) Having no training in geometry, I dream that the Pythagorean theorem is true. On that basis, and for no other reason, I come to believe that it is true. And of course it is. But it seems that the connection between the theorem and my dream that it obtains is arbitrary. To put it another way, I have no good reason to believe that the theorem is true. Or to put it yet another way, it appears that my belief is unjustified.



(2) I close my eyes, put my finger on the name of a horse on the racing form, and then bet the baby that the horse to whose name I have pointed to will win the fifth race. (The horse does indeed win the race.)



(3) While I sleep, I am anesthetized and whisked away to an operating room, where a mad scientist performs a surgical procedure on my brain. I am returned to my bed, and I awake to find myself disposed to assent to the utterance "Goldbach's conjecture is true." (Presumably some glitch has been inflicted on my "hard-wiring.") Goldbach's conjecture is that any even number greater than two is the sum of two primes. Let us suppose that Goldbach's conjecture is in fact true, but that no one has demonstrated it to be true.



(4) I come to believe on the basis of reading some tea leaves that the swallows have returned to Capistrano. Or take a case of an omen or a divination or an astrological forecast to the same effect. On the present account, such "occult" phenomena are in principle as capable of yielding knowledge as is the most careful observation. It need only be the case that the palm or the crystal ball or the tea leaves lead their interpreter to believe something that happens to be true. This case adds the dicomfiting element of ridicule. If the account I am giving really led us to put astrological predictions on the same footing as careful observation, the position would appear to be pernicious, its proponents credulous to the point of stupidity.



Now to claim that a position is stupid and pernicious is not to bring a philosophical objection to it. But the philosophical objection in all these cases is equally clear. The belief in question in each appears to be generated in a non-rational and unreliable way, and since knowledge is held to be rational, and so forth, these cases cannot be described as cases of knowledge.









II. Some Remarks on Belief


In case (3) as described we have, I agree, a deep-seated tendency to deny that the subject knows that Goldbach's conjecture is true, whatever the subject himself might claim. Nevertheless, the case itself is severely underdescribed. For example, am I supposed to know other propositions about number theory, which I bring to bear on my claim about Goldbach's conjecture? Do I, for example, know the difference between even and odd numbers, or what a prime number is? The strongest way to frame the counter-example, one which makes it clear that the belief is not justified on any account of justification, is to isolate the belief completely, to suppose that I am a mathematical naif. But a problem arises here, namely: is this a case of belief?


In the case as decribed I wind up, I know not how, with a disposition to assent, say to produce some word or conventional sign of agreement, when someone utters the sentence "Goldbach's conjecture is true" in a way that solicits my assent or dissent.(3) But if this is to be a case of belief, that cannot be the whole story, as we saw clearly in the first chapter. Such a disposition could be produced, for example, and as we have seen, in someone who spoke no English, and had no knowledge of mathematics, by a schedule of re-inforcement. (We starve the subject and then begin to feed him whenever he produces the desired response.) In a case such as this, it is surely misleading to say that the subject believes that Goldbach's conjecture is true. And in a case where I just started saying "Goldbach's conjecture is true" out of the blue, and I could not state the conjecture, or even say what a prime number is, the proper response would be "you said it, but you didn't even know what it meant: you didn't believe it." Again, as we remarked in the first chapter, the elements of the representation of a proposition have to function for the agent as referring to the proper items if the agent is to be said to believe the proposition under that representation. Something has to function for the believer as a representation of the proposition in question, or she simply cannot believe it. Here, that is going to require me to have considerable sophistication in mathematics. (Obviously, the causal theory of reference is going to be problematic with regard to numbers, since it is hard to see how a chain of causation can originate at an abstract object. But the point will hold good whatever the picture of reference in play in this case is.) That is, something is going to have to serve for me as a representation of Goldbach's conjecture. Nothing could so serve if I do not know what odd and even numbers, or prime numbers, are. So the believer cannot be a mathematical naif.


And consider also example (2). Here the relevant consideration is whether I have a sufficient degree of commitment to the proposition that the horse will win the race to be said correctly to believe it. Now, and again as we have seen in the first chapter, it is perfectly possible for me to act as though I believe something (say, by betting the baby on it) when in fact I do not believe it. If I am a compulsive gambler, I may look for some technique to pick horses without having any pronounced confidence that the technique is a good way to pick winners, or that any particular application of the technique will lead to the desired result. In such a case, I may act as though the proposition is true without believing it. And if the account I derived fom Newman in the first chapter is right, then it is not enough to believe that it is probable that the horse will win; that is not to believe that the horse will win. I have got to commit myself to the truth of that proposition. It is fair, I think, to say that people very rarely do that on the basis of their own guesses.


Thus, first, no belief stands in isolation; I cannot have the belief that Goldbach's conjecture is true and fail to have any related beliefs. The belief is constituted as a belief within a system of beliefs, and we could begin to give a systematic account of what this entails by seeing what is required in order for something to function for me as a representation of a particular proposition. In fact, the belief that Goldbach's conjecture is true depends not only on beliefs, but on knowledge.(4) Again, this should be explicated by appeal to conditions which must be met in order for something to function for an epistemic agent as the representation of a proposition. I must know some things about numbers, and I must know what Goldbach's conjecture is, in order to believe it. Further still, the claim that someone believes something entails that that person has some degree of serious and reflective commitment to the claim. If the claim is withdrawn as soon as it is brought into question, for example, we may rightly say that it was not believed. It is also surely possible to entertain or test hypotheses which we do not believe, or to treat them as if we believed them when in fact we do not.


So we need, on any account of knowledge, to have a sufficiently rich notion of belief to capture what is missing in cases such as these. That is, to formulate the case described in (3) in such a way that it counts as a counter-example to the view that knowledge is merely true belief, we must give a more substantive sense to belief than mere disposition to assent, and in case (2) we must give a more substantive sense to belief than a disposition to act in other ways as though a proposition is true. However, I need a sufficiently impoverished notion of belief to keep my view from collapsing into triviality, as it would do if I claimed that anything that counted as a belief was ipso facto justified. I think that (D) satisfies these constraints, although the present sort of point could stand even if that particular account were to be abandoned. Counting (2) and (3) as cases of knowledge will seem somewhat (though only somewhat) less implausible if they are framed in terms of belief rather than mere disposition to assent or to act as though committed to the truth of some proposition.


Arguments to the effect that some third condition is required for knowledge often play on an insufficiently rich notion of belief. Such arguments, again, often take the form simply of pointing out that a lucky guess does not count as knowledge.(5) But of course, in the usual case, a lucky guess is not even a belief. If I know that I am merely guessing, I would have to be epistemically perverse actually to believe, for example, that the horse to whose name I have pointed will win; at a minimum I would need some further beliefs about the reliability of my guesses. So it is really not enough to establish that some third condition is required that a lucky guess does not count as knowledge.



III. "How Do You Know?"


Another supposed fact about the use of the word `know' - also on the level of linguistic practice - that has been taken to lend support to the claim that knowledge is at least justified true belief, is that it is always legitimate, when someone claims to know something, to ask how she knows it.(6) I think that we have seen that in fact this is false - wildly false. In many cases of simple empirical knowledge-claims, such as the claim to know that here is a hand, or that I live in Oxford, the demand to produce reasons is completely misplaced. There are many, many instances in which it is completely inappropriate to press the question, in which the usual response to having it pressed would be incredulity, re-iteration of the claim, and so forth. So this argument, at least, if it is construed as an appeal to practice, cannot even begin. However, the claim might be that it is always, say, conceptually legitimate to demand "how do you know?" I am not completely clear about what conceptual legitimacy could mean here, but it should be clear that an appeal to conceptual legitimacy at this point would simply beg the question. That is, to say that it is always conceptually legitimate to press for a justification seems to amount to no more or less than the claim that justification is logically required for knowledge. So the assertion that such a demand is always conceptually legitimate could hardly be used in support of the conclusion that justification is logically required for knowledge.


But even if we allowed that the claim that it is always legitimate to ask "How do you know?" were true, and waived the fact that it begs the question, it ought to be be noted that this would not in itself establish a disanalogy between knowledge and mere true belief, or even between knowledge and mere belief, or even between knowledge and mere assertion. It is equally legitimate to ask someone who reports that she believes something why she believes it (here, the syntax of the terms makes the question of how she believes it rather different). Similarly, she might be asked "What makes you think so?" or even "Why do you say that?". In all these cases, justification may be demanded (though the request may also be for the causal genesis of the belief). (Though I would like to repeat: it is easy to overestimate the number of cases in which such demands are legitimate in practice.)


However, there is a disanalogy between the case of knowledge and the case of belief or assertion: in the case of knowledge, the answer to the question "How do you know?" may be relevant to whether in fact you do know. That is, if the question is not answered in a satisfactory way, it may be proper to respond that "You didn't know after all." In the case of belief or assertion the response to a demand for justification does not always bear on whether one does believe or assert what one says one believes or what one asserts. But again here, no disanalogy is established between knowledge and true belief, if the demand for a justification is thought of as a demand that the proposition which one claims to know be supported, that is, that reasons should be given to regard it as true. Eventually, I will argue elaborately that that is precisely how the demand for justification should normally be construed.


Be this as it may, the demand "How do you know?" can be pressed to the point of absurdity. If you tell a child that you know that elephants have trunks, the child may ask how you know it. Perhaps you tell her that you have seen pictures of elephants, or actual elephants, and that they had trunks. The child then asks how you know that what was pictured or seen was really an elephant. Here perhaps you appeal to authority, and by inference to a causal connection of some sort to the facts of the matter. The child then asks how you know the authority was telling the truth, and you may proceed by enumerating the authority's degrees, or by appealing to the authority's well-known veracity, implying that appealing to the authority is a reliable procedure for generating beliefs. The child may still not be satisfied. It is a revealing fact about the use of the word `know' that at some point you will silence your young interlocutor by saying that you just do know that some claim in the chain is true. That is, at some point you will assert that further justification is unneccesary. (In fact, under normal circumstances, you will reach this point much earlier than I have represented you as doing.) I am not pressing a regress here, but simply pointing out that the demand for justification will eventually be rejected with regard to some proposition embedded in the justification of the original claim. So (and as we already have very good reasons to believe) in some cases where we make a knowledge-claim, and where the believed proposition is in no sense self-justified or prima facie justified, we reject the demand for justification. And though this rejection may ultimately be illegitimate for conceptual reasons, it certainly does show that the actual use of the word `know' is very far from clearly suggesting that every case of knowledge is a case of justified belief.


On this basis I would like briefly to offer some counter-examples to the claim that knowledge is justified true belief. They are cases where it seems legitimate to reject the demand for a justification, cases where the demand seems inappropriate. We generate such cases where we are pressed to continue justifying claims which justify our knowledge, or where scepticism is pushed beyond tolerable limits. But there are other sorts of cases as well.



(a) As in the `elephant' case above, we often reject the last in a sequence of demands for a justification by claiming that no further justification is needed (and again, I am not thinking of cases of incorrigible reports about my own self-presenting internal states, and so forth).



(b) If the sort of points I have been exploring in the last two chapters are correct, then typical cases of knowledge about simple empirical states of affairs are counter-examples here. That is, it is quite within the bounds of practical propriety to laugh off a demand to say how you know where you live or that there's a pig under your nose. Some of the examples I am offering will seem fairly unusual, and it might not be thought that we should countenance as global a response to them as dropping any third condition on knowledge. But "Here is hand," "There is a pig," I live in Oxford" show, I think, that various strategies to finesse the unusual examples will not issue in a sartisfactory theory of knowledge.





(c) Some problems in a variety of disciplines have been solved in what is sometimes termed "a flash of insight," as when a mathematician who has been struggling with a problem suddenly "sees" the solution. Here, it might be argued that there is indeed a subterranean justification, for example, that ratiocinative processes are going forward at a sub-conscious level; after all, the mathematician has, by hypothesis, been working on the problem. Yet the proof might follow rather than precede the insight, and the mathematician might report that she knew the solution very well before she proved it. That is, she might report that she knew the solution but that a justification of it was not available to her. If we were to recognize the legitimacy of such a claim, we would have to admit that not all knowledge is justified true belief. Such examples are very far from being mere academic possibilities.(7)



(d) Here is another interesting case: that of religious faith. Let us for the sake of argument consider religious faith to be belief that God exists that is indifferent to any argument or objective evidence on either side of the question. (I assessed some of the literature on this topic in chapter two.) Then if we countenance the claim that religious faith could constitute knowledge if indeed God exists, we at once admit that not all knowledge is justified belief. And again, it is hard to see what makes this a case of knowledge (if it is) except that it is a case of true belief. I said at the beginning that no philosopher has explicitly held that knowledge is merely true belief. But any philosopher who holds that religious faith in this sense can rise to the level of knowledge is obliged to hold that not all knowledge is justified belief. (James may be a case in point, though Kierkegaard is not.) Or think of more mundane cases of faith, such as the faith we occasionally repose in persons. A father whose son is accused of murder might believe, in the face of overwhelming evidence, that his son is innocent. Asked how he knows, he responds impatiently that he just does. Let us stipulate that his son is indeed innocent. If this is a case of knowledge (as the father most assuredly believes) then not all knowledge is justified belief. And again, it is hard to see what makes this a case of knowledge (if it is), except that it is a case of true belief.



(e) Here is another sort of case in which we might recognize that someone has knowledge without justification. Consider circumstances in which it is appropriate to produce the utterance "I knew it all along." On some occasions, this means that though I believed something during a certain segment of time t (say, that Nixon was involved in the Watergate cover-up) a justification of it was not available to me at t. At t1 I come up with a justification for my belief. Then I might say not that I now know it, but that I knew it all along. In this case, though my belief comes at t1 to be justified, it is perfectly natural to say I knew it at t. This seems to me to mean that at t I believed it, and at t it was true.



Of course, even if we admit such cases, they would establish only that justification is not a necessary condition of knowledge; more would need to be said in order to do more than merely suggest that true belief is sufficient. But again, I am not at the moment attempting a demonstration, but offering a plea for the suspension of burden of proof. Nevertheless, even in cases such as these, we don't simply find ourselves believing something without any history or cause of our coming to believe it. For example, in the case of the flash of insight there is a process of inquiry in the context of which the belief is generated, though we are supposing that nothing in that inquiry directs one to the solution one proposes. In the case of religious faith, the belief is created and sustained by an emotional commitment to the truth of the believed proposition. (Indeed, if the view put forward in chapter one is right, all beliefs are created and sustained this way.) But if these sorts of processes count as justifications on the view of a proponent of the epistemology of justification, it is hard to see what beliefs are not justified. If the father is justified in believing his son innocent in the face of overwhelming evidence, then the notion of justification has been weakened to such a great extent that its continuing relevance to the theory of knowledge is questionable. For example, if the notion of justification is weakened to this extent, an account is required of why (1)-(4) do not count as cases of justified true belief (if they are described in such a way that they count as cases of belief).


And notice that if we admit this case, it will cut equally against internalist and externalist pictures of justification. For example, suppose that we are reliabilists; we hold that a belief is known if it is true and reached by a belief-forming mechanism that yields true beliefs most of the time. Surely, we cannot account for this case, because believing in the teeth of the evidence could not be a reliable belief-forming mechanism. Indeed, I suppose that an externalist could be fairly sanguine about the whole of the preceeding discussion (and I harbor no illusions that the externalist will cease to be sanguine after seeing the examples). A reliabilist can account for the intuitive mathematician (as long as her guesses are right more often than not, or if some particular answer is the result of a mechanism with a reliable propensity). And the reliabilist is probably in far better shape than the internalist with regard to simple judgements of perception. Indeed, for just that reason, I think reliabilism is far more promising as an overall account of justification than is internalism. But the cases of faith are far more difficult. Consider religious faith, for example. Here we are dealing with cosmological beliefs, beliefs about what the universe as a whole is like. As Hume might have argued, it is awfully hard to see which processes for generating beliefs about what universes as a whole are like could possibly be reliable, since we have only one universe to observe. If reliabilism is right, then I think no one does know any cosmological proposition to be true. Now again, I must admit that the damage of that fact to reliabilism is marginal, since many people in fact would deny that such propositions can be known. I think they can be, but I do not know how to mount a positive argument for that view. In any case, I will come back to externalism later, and give what I think are principled reasons why it ought to be rejected.



IV. Epistemological Anarchy


It may well be asked, however, how the present account is going to deal with cases such as those with which I started this paper. Now I suspect that the proponent of the view that knowledge is at least justified true belief is simply going to deny that cases like (a)-(e) are cases of knowledge. Likewise, I am simply going to affirm that cases like (1)-(4) are cases of knowledge, if they are described so that they are cases of belief. Notice, however, that, if someone claims to know a proposition that I do not believe to be true, or about which I have no opinion, it may be perfectly legitimate for me to deny the knowledge-claim when the claimant cannot produce good reasons. That is, it may be perfectly legitimate to deny that the claim is true, or at any rate to suspend belief about it, until good reasons to suppose that it is true have been produced. And in other cases it may be legitimate for me to deny that the claim is believed.


One way to put my point is that justification is a criterion, though not a logically necessary condition, of knowledge. Let us take a criterion, roughly, to be a test of whether an item has some property, a test that we apply if we are in doubt as to whether the item has that property or not. For example, it is a criterion for something to be gold that it yields a certain characteristic taste when bitten. In cases where we are in doubt about whether something is gold or not, we may well employ this criterion in deciding the matter. But it is hardly a logically necessary condition of something's being gold that it yields this taste when bitten. To see this, notice that, in a possible world in which the taste apparatus of persons is differently configured, gold would not yield the taste it does yield in this world. By contrast, the fact that an item has the atomic number 79 is a logically necessary condition for it to count as gold. Again, I claim that justification is a criterion for knowledge in the sense that, if the case is doubtful, the request for a justification acts as a test of whether S knows that p. But justification is not a logically necessary condition of knowledge.


Nevertheless the position that (1)-(4) count as cases of knowledge if they count as cases of belief sounds like rank irrationalism, like an argument for claiming to know anything you please on any grounds whatever, or on no grounds at all. It sounds like a defense of mysticism and obscurantism, of irresponsibility in inquiry, of charlatanry and quackery of all kinds. It seems that, on the present account, there is nothing wrong with coming to believe something on the basis of dreams, guesses, divinations. In fact, however, this charge is misplaced. There is something clearly wrong with such belief-generating strategies on the present account: they yield beliefs that are not justified.


It must be asked, however, just what the activity of justification is. On the account I will develop later, the practice of justification fundamentally consists in the attempt to ascertain or confirm whether some proposition is true. Now there is wide agreement among proponents of foundationalist, reliabilist, and even coherentist accounts of justification that justification must be truth-conducive, that is, that there must be reason to think that beliefs that are justified on any adequate account must be likely to be true.(8) (I am going to come back to this point at length.) This indicates, though it hardly proves, that justification is subordinate to truth, that our epistemic goal is true belief, while justification is a means by which we reach this goal and a means by which we confirm that this goal has been reached.


My opponents and I would agree that knowledge is a very valuable thing. Here, that means that, in the case of every proposition with which we are epistemically concerned, we ought to believe it just if it is true. On the present account, any process by which we come to believe the truth is a process by which we gain knowledge. Now careful argumentation and empirical observation are far superior to dreams, guesses, or divinations in this regard. So on the present account there is every reason in the world to pursue the former and neglect the latter. Hence, the claim that knowledge is merely true belief is hardly a prescription for epistemological anarchy.


Think of the circumstances in which it is in fact appropriate to press the question of how someone knows something. First of all, such a query is not usually made in circumstances in which the belief is not controversial or obscure. If someone claims to know that 2+2=4 or that the sky is blue on a sunny day, it would be odd, to say the least, in typical circumstances, to ask how she knows it. But if she claims to know that Goldbach's conjecture is true, or that there are hippopotami in Madagascar, or that Quayle will be the next president, then it may be appropriate to ask how she knows; that is, the criterion of justification comes into play. Notice that in the usual case, if I already believe that there are hippopotami in Madagascar, I will not press the query. This indicates that, in the usual case, I am trying to ascertain whether the claim is true by asking for a justification. But if I do not know whether the claim is true, I may well ask "How do you know?" or, with a shift in emphasis, "How do you know?". "Professor Ersatz gave a proof in his recent article"; "Hippopotami are indigenous to sub-tropical climes"; "The Democrats have no viable contenders." These are replies to the first sort of question. The second asks more specifically about the claimant's access to the truth of the matter. "Ersatz demonstrated it to me in the most convincing way"; "I just read an article in National Geographic"; "None of the touted candidates looks very formidable to me." In either case, the process of giving a justification can be plausibly construed as giving reasons to believe the claim is true. On the present view, then, such responses are indeed attempts to establish knowledge-claims. But the demand for a justification operates as a pragmatic rather than as a conceptual restraint. That is, justification is a practice that has as its goal to show that the conceptual conditions of knowledge are met.


On the other hand, if we did (as we occasionally do) ask how someone who claims to have knowledge of some strikingly non-controversial claim knows that the claim is true, if for example we ask someone how she knows that 2+2=4, we may have several different (but highly unusual) things in mind. For example, we may be pressing sceptical doubts about mathematics, in which case, again, we are concerned to establish that the believed claim is indeed true in the face of some considerations that make the other way. We might say in such a case that if the sceptical doubts prove to be sufficiently compelling, she did not know it after all. But again, what is meant is that there are good reasons not to regard the claim as true. On the other hand, and this is where the present account runs into difficulties, we may be pressing the question of the source of the belief. For example, if we find out that the claimant in this case has recently emerged from a mental hospital, and regards the voices in her head as reliable sources of information, we may well ask how she knows that 2+2=4. If she now replies that one of these voices told her, we may say (though with some strain to common sense) that she didn't know it after all.


Now my account indeed obliges me to deny this claim. The mental patient in this case, on my account, does indeed know that 2+2=4, if she satisfies the conditions for believing the claim. What we are doing in such a case, on the present view, is reaching for a truth by means of a literal falsehood. The truth we are reaching for is that knowledge-claims made by the mental patient on controversial matters ought to be regarded as highly suspect. That is, we are impugning the doxastic procedures of the claimant, because we have noted that the doxastic procedures she employs are likely in many cases to lead her to false beliefs. So as soon as she moves afield from pedestrian assertions that we all know to be true, her knowledge-claims ought to be subjected to scrutiny. As David Armstrong puts it, "the man who has mere true belief is unreliable. He was right this time, but if the same situation crops up again he is likely to get it wrong. . . . But if it is empirically impossible or even very unlikely that the situation will crop up again, then the distinction [between true belief and knowledge] loses almost all its point."(9) But Armstrong does recognize that the point of justification is to establish the truth of beliefs over the long haul. And it is worth noting (as Plato noted in the Meno) that the man who has true belief is perfectly reliable as long as he continues to have true belief. This should have made Armstrong suspect that the distinction between knowledge and true belief does lose its point. When we impugn someone's doxastic procedures, we are claiming precisely that he will not in fact continue by their application to generate true beliefs. Nevertheless, my view obviously obliges me to affirm that the case of the mathematically inclined mental patient is a case of knowledge. And notice that it is natural in a case such as this one to say that we all know that 2+2=4; it is "common knowledge"; in a typical case it would be perverse to ask of any one person how she knows it.


When someone claims to know something, and we ask how she knows it, when we demand a justification, we may be doing one of two things. First, we may be attempting to establish whether she does know it, or rather only believes it. That is to say that we may be trying to establish whether the believed proposition is true. Second, we may be trying to ascertain the believer's overall rationality, to ascertain whether she believes only what she has good reasons to believe. This will in turn affect our assessment of her further claims to know, which will be assessed in the same way, and so forth.


Thus, I am in no sense advocating the end of systematic epistemology. The view that knowledge is merely true belief neither makes the question of justification a trivial one, nor relieves us of the epistemological burden of producing an account of justification. Now the truth of propositions is rarely a matter of the noetic states of the believer. (The truth of the proposition that I believe that there are hippopotami in Madagascar is a matter of my noetic states, but the truth of the proposition that there are hippopotami in Madagascar is a matter of the way the world is.) In establishing a knowledge-claim per se we are only concerned about the truth of the believed proposition; the view that knowledge is merely true belief makes the relation of a belief to other of the believer's noetic states strictly speaking irrelevant to whether he has knowledge, except insofar as such relations are required for belief, and leaving aside the case of belief about the believer's internal states. Nevertheless, we have also seen that the demand for justification may be a way of impugning or at any rate querying the doxastic procedures of the believer. The difference here is parallel in some respects to the difference between questioning the rightness of a person's actions and questioning the goodness of the person herself. It is like the difference, that is to say, between asking whether she does the right thing, and asking whether she does the right things for the right reasons. Here, we do demand that the person who makes a knowledge-claim tell us something about the relation of her belief to other of her noetic states (as well, perhaps, as to external matters). For example, we may be trying to determine whether the claimant is generally thoughtful, whether, as it were, she makes her beliefs run the gamut of her other beliefs and is at pains to be consistent. We might, that is, perfectly well be trying to determine whether some holistic coherence relation obtains between this belief and others. And note that, if the proposition she claims to know is controversial, our overall judgement of the claimant's rationality may well affect our assessment of the proposition's truth, and hence of the truth of the knowledge-claim itself.



V. Lycan's Ad Hominem Argument


There is a final objection I would like to consider. William Lycan (in correspondence), has put forward what he calls the "ad hominem argument." He abbreviates the claim that knowledge is merely true belief `TB,' and the argument runs as follows:



(1) Sartwell Believes TB. [Premise; I assume he is sincere.]


(2) TB is true. [Assumption for reductio.]


(3) If TB is true, then whoever believes truly that p knows that p. [Trivial, given what TB says.]


(4) If (1) and (2), Sartwell knows TB. [From (1), (2), (3).]


(5) Sartwell knows TB. [(1), (2), (4).]


(6) For any subject S, if S believes that p and S is well aware that S believes that p, then S believes that S believes p truly. [Premise.]


(7) Sartwell is well aware that he believes TB. [Premise; I assume he is reflective as well as sincere.]





(8) Sartwell believes that he believes TB truly. [(1), (6), (7); also straightforward on its own.]


(9) For any subject S, if S believes p and S knows that q is equivalent to p by philosophical analysis, then S believes q. [Substitutivity of known analytic equivalents in belief contexts.]


(10) Sartwell believes he knows TB. [(2), (8), (9).]



This is bad enough, as Lycan points out, given that TB is a controversial philosophical thesis, but it gets worse:



(11) For any p such that Sartwell both believes p and is well aware that he believes p, Sartwell believes he knows that p. [(2), (6), (9).]



Lycan wites: "(11) rules out Sartwell having any self-consciously modest belief whatever, i.e. any belief which he is aware of holding but of which he is not confident enough to make a knowledge claim. I do not know him personally, but I cannot imagine that in real life Sartwell is so opinionated."


Now I am an opinionated little chump, and I do embrace (10), at least most days. But (11) strikes me as extremely disturbing. And indeed, I like many other people sometimes say: "I believe it, but I don't know it." That sounds like a contradiction for one who holds my view. One possible place to drive a wedge would be (6), which looks to me like the only controversial line of the derivation. But notice, by the lights of the first chapter, I am not only committed to (6); I am, with some qualifications which I will not enter here, committed to the claim that (6) is analytic. Nevertheless, I think that the view sketched in chapter one is capable of taking some of the sting out of the ad hominem argument, even while admitting that it is sound.


Recall that I there took up Newman's account of degrees of belief. To believe p tentatively, I held, is not to believe p at all, but to believe that, say, p is more probable than its competitors. That is, one does not, in such a case, take some `p' to express a truth, but rather, one takes some `p is probable' to express a truth. Loosely speaking, as an abbreviatory device, we refer to the latter as believing p conditionally, modestly, and so forth.


This account allows me to have my arrogance and eat it, too. In the sort of case envisioned in the argument, which might initially be described as my tentatively believing some proposition, there is a proposition that I believe with no tentativeness. And for that proposition, the ad hominem argument goes all the way through. So I can be modest about p and immodest about "p is more probable than its competitors." As to the case of my own belief in TB: I get worked up enough at times to enthusiastically embrace (10); I do think I know that TB is true. At other times, I entertain some lingering doubts. But I am at least always willing to endorse the claim that TB is more probable than its competitors.


This still leaves me with the task of accounting for such locutions as "I believe it, but I don't know it." Note that this locution does make sense in the mouth of someone who believes that knowledge requires justification, but it seems to make no sense in my mouth. Nevertheless, it sometimes emerges therefrom. Now I would like to make an ordinary language point about this locution: "I know p" is often used to express a great deal of confidence that p. This usage makes perfect sense if knowledge is mere true belief. To say "I know p" is often just to say, with special emphasis, "p is true, true, true." That is, one is saying that p is not merely a belief, it is a true belief. And note, this use of the term `know' does not lend comfort to someone who thinks that knowledge requires justification. It is a way of expressing confidence, psychological certainty. But psychological certainty is not justification. In the mouth of a believer, "I don't just believe that God exists, I know it," is just as often an expression of faith as it is an assertion of justification. So on the present view, to say "I believe it, but I don't know it," would be to withold this special emphasis.


VI. Internalism and Externalism


It may well be asked what substantive advantages accrue for epistemology from accepting the view that knowledge is merely true belief. Obviously, my account yields an economical definition of knowledge. That, however, is of no help if the account is not itself plausible. But further, even if the account is not plausible, it may serve to challenge philsoophers who take justification to be a logically necessary condition of knowledge to defend their view. I am unacquainted with any argument in the literature to the effect that justification is a logically necessary condition that is not fundamentally an appeal to intuition. It is worth noting that such intuitions (intuitions which, for example, motivate the Gettier problem) are not universally shared, and thus require defense.


A more fundamental advantage of the view that knowledge is merely true belief is that, on it, we are under no apparent pressure to choose between a broadly externalist and a broadly internalist account of justification. The pressure to choose between these views arises largely because proponents of each argue that their account is an account of the sense of justification that is logically required for knowledge. But justification in no sense is logically required for knowledge.


Externalists make justification fundamentally a relation between a belief and the way the world is. For example, the justification of a belief is construed as a matter of the objective reliablility of the method by which the belief was generated(10), or of a lawlike relation between the belief and the state of affairs with which it is concerned.(11) A major problem for any such account is that it is compatible with a belief's being justified that the believer is in some sense irrational in adopting it, as would be the case if the believer did not believe that the belief was arrived at in a reliable way, or that the proper nomological relation held. Internalism, on the other hand, makes justification a matter of the relation between the belief in question and certain other of the believer's noetic states. For example, the question of justification has been construed as being a matter of its inferential relation to basic beliefs or non-doxastic introspectible states,(12) or as a matter of holistic coherence within the overall doxastic structure.(13) A problem here is that such an account stands in need of what BonJour calls an external "metajustification," an argument to show that the whole doxastic structure is not fantastic, is somehow tied to the world and hence that beliefs that have the proper position in the structure are likely to be true. Each broad kind of account does an admirable job of filling the deficiencies of the other. That is, externally justified beliefs are by definition likely to be true, while internally justified beliefs are by definition likely to be rationally adopted.


Now on the present account of knowledge, I can see no reason whatever to choose between these alternatives. As we have seen, on some occasions where a justification is demanded, the demand is for reasons to regard the believed proposition as true, whereas on other occasions, the demand is for reasons to regard the believer as rational. Since justification is not part of the theory of knowledge, both projects ought to be pursued: the externalist ought to continue trying to show how we ought to go about generating true beliefs, while the internalist ought to continue trying to show how those beliefs should be incorporated in a doxastic structure in order for the structure to be rational. One might think of this as the contrast between showing that a belief is justified and showing that some person is justified in adopting it. But the debate now ought to proceed among externalists and among internalists rather than between them.


So I think that the view that knowledge is true belief is capable of doing a service to epistemology in the sense of cogently addressing some of its outstanding problems. But the most conspicuous advantage of the present over the traditional view is that it can allow that there is knowledge that is not constructed with the use of and cannot be manipulated with the tools provided by theories of justification. This point, I think, follows from the discussions of scepticism and of veridical experience in the last two chapters. Problems are solved in a variety of ways, ranging from the reliable and rational to the perfectly inexplicable.(14) We should not falsify our own intellectual lives in a philosophical reconstruction of how those lives are conducted by pretending that knowledge always proceeds along the orderly path of justification.


Now that we have answered some of the obvious objections to the claim that knowledge is merely true belief, and now that we have enumerated some of the advantages that accrue if we accept this view, we can go on to mount a positive argument for the analysis.



VII. Terminological Groundwork


One objection that has been raised to the claim that knowledge is merely true belief is that it seems to leave us without any pre-analytic sense of what knowledge is. That is, it seems to be a question whether the proffered analysis could possibly be an analysis of knowledge at all.(15) This objection seems to arise because there is (supposedly) a very widely, if not universally held, intuition that belief and truth could not possibly be sufficient for knowledge. Then the question of whether the analysis embodies a theory of knowledge seems to come down to a clash of intuitions about what notion it is that is being analyzed. For this reason, I want to offer a way of picking out the notion to be defined. Here is my proposal: knowledge is our epistemic goal in the generation of particular propositional beliefs. It will be convenient to have a term to refer to procedures that have as their goal the generation of belief with regard to particular propositions. Let us refer to such procedures collectively as inquiry. (This use of the term `inquiry' is to be regarded as a sheer stipulation.) Knowledge, then, is the purpose or telos of inquiry. (This minimal conception of knowledge is, I believe, in keeping with the traditional use of the concept of knowledge for instance, in Plato, Descartes, and Hume.)


I say that inquiry comprises procedures that have as their goal the generation of belief in particular propositions, because there are other, wider-ranging epistemic goals besides knowledge: for example, rationality and wisdom. But such concepts take beliefs in groups or even consider the entire structure of beliefs of an epistemic agent. It can be rational or wise, as part of such an overall structure, to believe that p. But one cannot have the rationality or wisdom that p, as one can have the knowledge that p. At any rate, we might think of a theory of epistemic justification as doing two things: (a) giving general procedures for inquiry, and (b) evaluating the productions of inquiry, that is, particular propositional beliefs. A belief, then, is justified if it is produced (or, alternately, could have been produced) by a correct procedure for inquiry(16), and to say that a belief is justified is to evaluate it positively along epistemic lines. I regard both projects as worthwhile. I am, again, by no means arguing that an account of justification is unimportant; my claim, rather, is that such an account is not part of the theory of knowledge.


To continue with terminological groundwork, it is possible to distinguish two sorts of epistemological projects: normative and descriptive. Descriptive epistemology is essentially a branch of psychology or cognitive science. It asks: How do we come to believe or to know? What cognitive mechanisms and causal circumstances operate in the acquisition of belief or knowledge? What, e.g., is the function with regard to belief and other propositional attitudes of perception, memory, and so forth? Normative epistemology, on the other hand, is what has usually been thought of as a distinctively philosophical discipline. It asks: How should we generate beliefs with regard to achieving knowledge? How should inquiry be conducted and evaluated? And what, exactly, is its goal, i.e., what is knowledge?


Now these two sorts of investigations are by no means unconnected.(17) For example, it would be ludicrous to recommend (normatively) a procedure for generating beliefs that was literally impossible (descriptively) to apply. Thus, the normative project might be thought of as giving guidelines for the correct or optimal application of the procedures we do use to generate beliefs. And if the basic emphasis of this book is correct - if it is true that epistemology is concerned with the norms by which we in fact conduct ourselves intellectually, if the basic question of epistemology is "What ought we to believe?" - then normative epistemology must be constrained by descriptive epistemology in the way suggested. At any rate, I will be concerned in the remainder of this chapter with normative epistemology.





VIII. Deontological and Teleological Conceptions


The most elaborately developed normative theories are in ethics, and thus normative epistemology often relies on a parallel to ethics. Ethical systems have been divided into two kinds: deontological and teleological.(18) Proponents of the former think of moral action as what is done in obedience to principles which serve in turn no end that could be looked on as an overall moral goal. Moral action is to specified in terms of obligation and permission. If I do only what is permissible (possibly, if I do it because it is permissible), or what is demanded by duty (possibly, if do it because duty demands it), then I am not subject to ethical disapprobation even if the result of my action is disastrous. According to proponents of teleological ethics, on the other hand, an action is morally good when it conduces to some goal, for example, the greatest happiness of the greatest number, or if it is in accordance with some rule the observance of which so conduces. Similarly, there might be two sorts of normative epistemology: one which prescribes duties and permissions in generating beliefs (and other propositional attitudes) without regard to any overarching epistemic goal, and one which prescribes some goal for epistemic activity, and recognizes the legitimacy of any procedure that conduces to that goal, or, alternately, of any procedure which accords with certain rules the observance of which in turn conduces to that goal. I have asserted that knowledge is the goal of inquiry. But this supposes that inquiry has some goal, which would be denied by a proponent of deontological normative epistemology. So we had better start with a discussion of whether that position is plausible.


The taxonomy of normative epistemology suggested by this particular parallel to ethics has been developed by William Alston. Because my discussion follows his to some extent, I should pause here to diffentiate my use of terms from his. Alston uses the term `deontological' to distinguish systems which epistemically prescribe, proscribe, or permit certain beliefs or belief-generating procedures, from what he terms `evaluative' systems, which merely assess certain beliefs and procedures from the standpoint of some standard.(19) He points out that it is not the case that all standards of evaluation depend on such concepts as obligation and permission, that not all standards carry with them the implication that the subject is praiseworthy for meeting them or blameworthy for violating them. For example, to say that some person is beautiful is to evaluate her appearance positively, but it is not to say that she is praiseworthy for her appearance, since she may not be responsible for it; it may be a genetic endowment.(20) The relevant point here is that both sorts of systems (Alston's `deontological' and `evaluative') are what I term `teleological'; he describes both as being directed to the goal of generating true belief and avoiding false belief.


It may be a question, then, whether any philosopher has seriously held a deontological position in my sense, has seriously held that we have some epistemic obligations but that there is no overarching goal of inquiry. Some extreme idealists and postivists, who identify truth and justification, may harbour such a view. If one has a coherence theory of truth and also a coherence theory of justification, for example, then one may simply count as knowledge whatever beliefs are generated by whatever procedures turn out to embody justification; if it was supposed to be a sheer fact that we ought to follow such procedures, if there were no further goal in mind, this would be a deontological position in my sense. The notion of knowledge is in some sense superfluous on this position; at least, it does not describe a distinctive purpose for inquiry above the fulfillment of certain duties or obedience to certain rules. Clearer examples of deontological views could be proposed: for example, believe all and only the propositions contained in the Bible, or in the writings of Mao.


Deontological views in my sense have, these days, few proponents, and seem on the face of it extreme and implausible. Their implausibility can be brought out in the following way. What is the source of our epistemic obligations? Or, to put it another way, is there any good reason to think that we have any distictively epistemic obligations at all, in the absence of some overarching purpose for inquiry? The same problem arises for deontological moral theories, but here there are plausible, or at least fairly widely proffered, answers: our moral obligations derive from God, for example, or from the state. Again, it is possible that the very same sources yield our epistemic obligations. But to establish this, we would have to give good reasons to think that God does impose epistemic obligations, or to give an account of the "epistemic legitimacy" of the state. Furthermore, there no longer appears to be any distinction between moral and epistemic constraints on the generation of beliefs. There no longer appears to be any distinctively epistemological enterprise.


I will take it, until such objections can be answered, that a purely deontological conception of epistemic norms is highly implausible. Indeed, no contemporary philosopher with whose work I am acquainted holds such a view. It might be asked, then, why I have bothered to discuss it at all. The point is just this: inquiry does have a purpose; only teleological conceptions of normative epistemology are plausible. But what about deontological conceptions in Alston's sense? Such conceptions formulate obligations and permissions, and award praise and blame, with regard to the way these obligations and permissions conduce to an overarching epistemic goal. I will term such views, in keeping with my own rather than Alston's use of terms, injunctive teleological accounts of normative epistemology. They might be thought of by analogy to rule utilitarianism, where one's obligations are to follow rules the observance of which tends to the greatest happiness of the greatest number. Rule utilitarianism has the advantage over act utilitarianism that it can yield relatively detailed injunctions on conduct, rather than the huge amorphous injunction to act so as to create the greatest good of the greatest number, which might be extremely difficult to appply in individual cases. Likewise, injunctive teleological views yield duties or permissions which, if observed, will conduce to knowledge, rather than the huge amorphous duty to generate beliefs so as to gain knowledge.


Alston attacks injunctive teleological accounts as pre-supposing doxastic voluntarism, the view that our beliefs are, in one sense or another, under our voluntary control. After all, I cannot have an obligation to do something or to refrain from doing it if it is not up to me whether I do it or not. Injunctive teleological conceptions of justification indeed suppose doxastic voluntarism. And Alston's objections to doxastic voluntarism appear to me devastating. Doxastic voluntarism and the issues surrounding it were already broached in the first chapter, though there is obviously much more to be said about the matter than I said there. But for current purposes we need not enter very deeply into this ongoing debate. For a detailed discussion of the issue, I refer the reader to Alston.(21) Here, I recount a single example: when I see a tree before me, and have no reason to think that, for example, I am dreaming, or that the perceptual situation is grossly abnormal, or that someone is replacing trees in my neighborhood with plastic replicas, it is simply not up to me whether I believe there is a tree before me; like it or not, I will believe it. So I should hardly be praised or blamed for doing so. It should be noted that this criticism applies not only to injunctive teleological accounts but also to deontological accounts in my sense.


If this criticism is plausible, we are left with a conception of justification that Alston terms an "evaluative" or "strong position" account. These terms describe justification from the two angles mentioned above; to say that it is desirable to get into a strong position is to give a view about (a), the correct procedures for inquiry, while to evaluate someone's epistemic position positively (though not in deontological terms) is to apply (b), some account of normative evaluation with regard to these procedures. We ought not, on Alston's view, to alot praise and blame in epistemic matters, but some epistemic situations are better than others. (It is better to be beautiful than ugly, but, again, one is not usually praiseworthy or blameworthy in this regard.) On Alston's view, the most plausible way to formulate this conception is that one's belief is justified if it is based on adequate grounds, and having adequate grounds is a matter of being in a strong position to generate a true belief on the matter. It is a good thing to be in such a position, but one is not, or at least one is not always, responsible for getting into such a position. Now it is not part of my current project to defend this conception of justification (though I regard it as plausible); rather, I want to use it in a certain way to elucidate what, on it, would be a plausible view of our epistemic goal with regard to particular propositions, that is, what would be a plausible theory of propositional knowledge.



IX. Truth-Conduciveness


It is widely held that our epistemic goal with regard to particular propositions is achieving true beliefs and avoiding false ones about propositions with which we are epistemically concerned. (We have seen that Alston, for one, endorses that view.) That is, it is widely admitted that on any good account of justification, there must be reason to think that the beliefs justified on the account are likely to be true. Indeed, proponents of all the major conceptions of justification hold this position. For example, the foundationalist Paul Moser writes:



[E]pistemic justification is essentially related to the so-called cognitive goal of truth, insofar as an individual belief is epistemically justified only if it is appropriately directed toward the goal of truth. More specifically, on the present conception, one is epistemically justified in believing a proposition only if one has good reason to believe it is true.(22)





The reliabilist Alvin Goldman claims, similarly, that a condition on an account of justification is that beliefs justified on the account be likely to be true; he says that a plausible conception of justification will be "truth-linked."(23) And the coherentist Laurence BonJour puts it even more strongly:



If epistemic justification were not conducive to truth in this way, if finding epistemically justified beliefs did not substantially increase the likelihood of finding true ones, epistemic justification would be irrelevant to our main cognitive goal and of dubious worth. It is only if we have some reason to think that epistemic justification constitutes a path to truth that we as cognitive human beings have any motive for preferring epistemically justified beliefs to epistemically unjustified ones. Epistemic justification is therefore in the final analysis only an instrumental value, not an instrinsic one.(24)





In fact, it is often enough taken to be the distinguishing mark of the fact that we are epistemically concerned with a proposition that we are concerned with its truth or falsity. That is what, on the view of many philosophers, distinguishes epistemic from moral or prudential constraints on belief, what distinguishes inquiry from other belief-generating procedures. (If the theory I gave in the first chapter is right, ther are no non-epistemic belief-generating procedures in this sense. That fact merely underscores the present point.) I have argued that a plausible normative epistemology will be teleological. And I have claimed that the conception which accounts of knowledge are attempting to analyze or describe is that of the epistemic telos with regard to particular propositions. It would follow that, if a philosopher holds that the epistemic telos is merely true belief, that philosopher implicitly commits himself, his own asservations to the contrary, to the view that knowledge is merely true belief.


I think that this is the case. I think, that is, that in the above passages, these philosophers have committed themselves implicitly to the view that knowledge is merely true belief, and that justification is a criterion rather than a logically necessary condition of knowledge. By a criterion, to repeat, I mean a test for whether some item has some property that is not itself a logically necessary condition of that item's having that property. Justification on the present view is, first of all, a means by which we achieve knowledge, that is, by which we arrive at true beliefs, and second, it provides a test of whether someone has knowledge, that is, whether her beliefs are true. So again, the present view does not make accounts of justification trivial, or unconnected with the assessment of claims to know. If our epistemic goal with regard to particular propositions is true belief, then justification (a) gives procedures by which true beliefs are obtained, and (b) gives standards for evaluating the products of such procedures with regard to that goal. From the point of view of (a), justification prescribes techniques by which knowledge is gained. From the point of view of (b) it gives a criterion for knowledge. But in neither case does it describe a logically necessary condition for knowledge.


Another way of putting the matter is like this. If we describe justification as of merely instrumental value with regard to arriving at truth, as BonJour does explicitly, we can no longer maintain both that knowledge is the telos of inquiry and that justification is a necessary condition of knowledge. It is incoherent to build a specification of what are regarded merely as means of achieving some goal into the description of the goal itself; in such circumstances the goal can be described independently of the means. So if justification is demanded because it is instrumental to true belief, it cannot also be maintained that knowledge is justified true belief.


I will now certainly be accused of begging the question by assuming that knowledge is the goal of inquiry. There is justice in this claim in that I have not gone very far toward establishing the point. But I would ask my accusers at this point whether they can do better in describing the conception which theories of knowledge set out to analyze or describe without begging the question in favor of some such theory. And I ask also, if knowledge is not the overarching epistemic telos with regard to particular propositions, why such tremendous emphasis has been placed on the theory of knowledge in the history of philosophy, and just what function that notion serves within that history. If knowledge is not the overarching purpose of inquiry, then why is the notion important, and why should we continue to be concerned in normative epistemology above all with what knowledge is and how it can be achieved? If we want to withold the term `knowledge' from mere true belief, but also want to hold that mere true belief is the purpose of inquiry, then I suggest that what remains is a mere verbal dispute. That is, if we treat mere true belief as the purpose of inquiry, but do not equate it with knowledge, then I do not think that knowledge is any longer central to normative epistemology. And I would insist that we are not going to understand what `knowledge' means in the tradition, in Plato and Descartes, for example, if we do not regard them as holding knowledge to be the goal of inquiry. In fact, if it is allowed that mere true belief is the telos of inquiry, but that we should still reserve the term `knowledge' for justified true belief (and perhaps something more), I will simply abandon the term `knowledge' to the epistemology of justification. But first of all, as I suggested in the third chapter, I think that `knowledge' will now merely be a technical term with a stipulated definition. And second, I do not think it will be central to epistemology, since it no longer represents our epistemic goal. And third, I think the stipulated definition will either be redundant (if justification is held to be truth conducive) or, as I will argue, incoherent (if it is not).


Now it may well be held that justification is of more than instrumental value, because if we are not justified in believing p, though p is true and we in fact believe it, we may have false beliefs that lead us to p, and we may continue to generate false beliefs in the future. All of this is true, but it is irrelevant to the present point. Recall that I have characterized knowledge as our epistemic goal with regard to particular propositions. Insofar as p is concerned, this goal has been realized if p is true and we believe it. Insofar as we have also such goals as continuing to generate true beliefs, rendering our system of beliefs coherent, and so forth, it is desirable to have justified beliefs. But with regard to any particular proposition, our goal has been reached if we believe that proposition and it is true.


But I do not want simply to let the matter rest on a supposed agreement among some contemporary epistemologists that our epistemic goal with regard to particular propositions is true belief. Such epistemologists are agreed that knowledge is at least justified true belief. I think that Alston is right to think that the only plausible way to construe this claim is that knowledge is at least true belief based on adequate grounds, or true belief reached from a strong position. So perhaps the figures in question, on reflection, would describe the epistemic telos not as true belief but as true belief based on adequate grounds, or true belief reached from a strong position.


Only it must now be asked, why do we want to have adequate grounds? Why do we want to be in a strong position? This question ought to be misguided if true belief based on adequate grounds or true belief reached from a strong position is in fact the purpose of inquiry. For there is no good answer to the question of why we desire our ultimate ends. But the question is hardly misguided. In fact, we cannot even specify what it is to have adequate grounds except that these grounds tend to establish that the proposition in question is true; we cannot even specify what it is to be in a strong position except as being in a strong position to get the truth. This indicates that the purpose of inquiry can be formulated without reference to the notions of ground or position. Thus, on the views in question, believing the truth is in fact our overarching epistemic telos with regard to particular propositions, on the only plausible conception of justification. Hence, on these views, knowledge is merely true belief.



X. Externalism


Now I regard externalism with sympathy as a theory of justification. I believe that an agent can be justified in believing p in virtue of facts of which the agent is not aware. In particular, I think some version of reliabilism of the sort endorsed by Goldman is very compelling indeed. The reasons for my sympathy emerged in the third chapter; I think that internalist pictures of justification cannot count most very simple empirical beliefs as justified. But such beliefs are, I think, when true, paradigm cases of knowledge. Various kinds of internalism presuppose that there is an internal component of veridical experience which is necessary in order to have the experience in question and which is epistemically available to the doxastic agent. But, as I have argued, there are no good reasons to accept this claim, and several good reasons (drawn from practice) to reject it. So I regard the reliabilist program as compelling, particularly with regard to propositions such as "Here is a hand," and "I live in Oxford," since one could be a reliabilist and not accept internalism with regard to veridical experience.


For example, Goldman's view about justification is roughly this: a system of justificatory rules is right just if it permits certain psychological processes which result in the generation of true beliefs in some high ratio of circumstances.(25) At a minimum, such a ratio must be over .50. Provided we construe the notion of a "psychological process" widely enough so that such processes are not limited to the head of the agent, but could encompass fused relations of the agent to her environment, such a view is compatible with radical externalism with regard to veridical experience. As long as being in such fused relations was a reliable psychological process, and the system of rules suggested permitted such processes, Goldman would count that process as lending justification to the generated belief. (It would be better to reformulate Goldman's view to get away from the notion of "permission," which must be associated with a deontological or injunctive teleological conception of justification. It would be easy enough to substitute an "evaluative" conception in some such way as this: A system of justificatory rules is right if it yields the result that the products of reliable proceses are to be evaluated positively along epistemic lines.) So as a proponent of (ER), I am certainly motivated to prefer Goldman's reliabilism to internalism, especially with regard to simple empirical propositions.


But I think we have already seen reasons to believe that reliabilism could not possibly yield a necessary condition for knowledge. That is because reliabilism stakes everything on truth-conduciveness. If true belief is not our epistemic telos (and I will suggest, a bit later, that there are decisive reasons to think it is), then reliabilism would lose its raison d'etre. For on Goldman's view, reliability of a process is construed precisely as consisting in the fact that the process yields a high ratio of true over false beliefs. Evidently, any technique which yields a high ratio of true over false beliefs is capable of figuring in some system of correct justificatory rules. This is absolutely as it should be. That is, if reading tea leaves is in fact highly reliable, then it makes sense to consult tea leaves. (It is precisely because reading tea leaves is not reliable that "the tea leaves say so" is a very bad justification for any belief.)


But now we are in a position to see that reliabilism construed as part of the theory of knowledge falls victim very directly to the sort of objection I have been pushing. If a belief-generating technique is to be meta-justified by its truth-conduciveness, then the epistemic goal in which it is in service is mere true belief. Reliabilism builds a specification of the mere means of reaching a goal into the description of the goal itself. It follows that the goal can be described independently of that specification. And since knowledge is our goal with regard to the generation of particular propositional beliefs, it actually follows from reliabilism that knowledge is merely true belief. The very same argument, with slight variations, can be shown to apply to nomological externalism of the sort put forward by Armstrong, and to causal versions of externalism of the sort endorsed by Goldman in an earlier incarnation.



XI. The Race


To conclude the discussion of accounts of justification that require justification to be truth-conducive, I would like to consider an analogy that has been repeatedly made in refutation of the treatment I have suggested. Justification, the analogy goes, is like a race, and true belief is its finish line. Winning the race is achieving knowledge. But you do not win simply in virtue of crossing the finish line; you must run the reauired distance along the required route. That is, to win the race (achhieve knowledge), you must reach the finish line (true belief) in the right way (the way of justification). You may recall the case of Rosie Ruiz, who "won" the Boston Marathon by jumping in with a couple of miles to go. To identify knowledge with true belief is to say that Rosie should have been declared the winner, that any way to reach the finish line is acceptable.


However, consider the relation of means to ends in that case. If we ask: "why must one run the required distance along the required route?" the answer is simply that one must. Those are the rules of the race. We have reached the level of brute fact, the level of the external conditions that make the race the race it is. To participate in the race is to agree to abide by the rules, and the exact length of a marathon does not itself admit of any justification.


Now if we ask: "why must we reach true belief in the right way?" the analogy suggests an answer: we just must, those are the rules. Now this returns us directly into a deontological account of epistemic justification. I have already argued that such an account is implausible, and here we again see why it is. The account makes justification arbitrary; it simply prescribes justificaton as the rules of the race. But that justification is not arbitrary in this way can be seen in several ways. First, we could, for convenience, simply stipulate whatever rules we found fun, or convenient, or entertaining. But the thinkers in question do not think we can stipulate any rules we like. Second, it would be a bizarre exercise at this point to engage in a debate about justification; that would be like debating the general question "How long should a race be?" That question obviously has no correct answer: a race can be how ever long we stipulate it to be. A fifty yard dash is not more correct than a marathon. This would make the entire debate about epistemic justification absurd.


I would like to suggest another analogy, which I owe to Arthur Skidmore. Inquiry is like a journey, and true belief is its destination. Justification is a map. Now using the map will help us arrive at our destination. But the point is to arrive. If we misread the map and get there anyway, everything has turned out fine. If we take a route not marked on the map, we have still done as well as could do on this particular journey. I suggest, at a minimum, that an account such as this is compatible with the views of, say, Moser, BonJour, and Goldman.



XII. Lycan's Explanationism


The only fully developed, contemporary account of justification of which I am aware that appears to steer clear of this consequence (and indeed, one of the best contemporary treatments of justification) is that put forward by William Lycan in Judgement and Justification. Lycan proceeds by denying that truth is the only or the overall purpose on inquiry. His claim is that beliefs are justified in virtue of their explanatory role; they are justified by their coherence with previously held beliefs, by their simplicity, their scope, and so forth. And Lycan even argues that all inference relies on a (perhaps implicit) appeal to best explanation. For example, even in a simple modus ponens deduction, we are not forced to accept the conclusion; we always have the option of rejecting one of the premises. So Lycan concludes that all reasoning involves comparisons of plausibility. He describes justified beliefs as those which increase the overall explanatory coherence of one's system of beliefs. Now Lycan's position is in some respects similar to BonJour's; he holds that we have "spontaneous," that is, non-inferred beliefs that have some degree of prima facie justification (due to the explanatory virtue of conservatism, which states that we should not without some explanatory motivation change the beliefs we already do have). But Lycan's position differs in two crucial respects from BonJour's. First, Lycan goes further toward elucidating the notion of coherence in his description of the explanatory virtues. Second, though, as we have seen, BonJour holds that any coherence theorist must give a "meta-justification" of his view, that is, he must give reasons to think that beliefs that are justified on the view are likely to be true, Lycan demands no such thing.


In fact, Lycan goes so far as to say:



[I]f . . . appeal to simplicity [e.g.] is a fundamental epistemic method, then it is fundamental; no further question can arise regarding its "connecton to truth." . . . For that matter, I do not in the first place accept the a priori assumption that epistemic justification must be a matter of quantitative relation to truth.(26)



Lycan does indeed give a "meta-justification" of his account, however. Simplicity, power, and the rest are useful on the evolutionary scale. They are adaptive. I am not sure how plausible this is, but the point is that, if we are to hold that justification is a component of knowledge, we need to take Lycan's general approach. That is, we must describe justification as conducing to some goal other than truth, or else lay down justification as a deontological desideratum. We cannot, as we have seen, coherently maintain that knowledge is the telos of inquiry, that justification is a component of knowledge, and that accounts of justification must in turn be meta-justified by their conduciveness to truth. And Lycan manages to avoid also the dangers of a deontological epistemology; he does not in fact regard simplicity or power as sheer desiderata, but as useful adaptations. Nor does his view assume doxastic voluntarism.


Nevertheless, the view has severe drawbacks. To begin with, it makes the distinction between epistemic and other values obscure. In fact, it absorbs epistemic value into sheer utility, and it is not at all clear that this is a plausible line. It is true, as Lycan admits, that there is no a priori reason to think that a simple hypothesis is more likely to be true than an extremely complex one. But a simple hypothesis has less cognitive cost; it is cognitively efficient. That may make it seem that it is better in the long run to believe simple hypotheses. But it is not clear that it is in fact, better, epistemically speaking, to believe simple hypotheses.


To get clear on the point, let us suppose (as seems likely, though Lycan is reticent here) that Lycan holds justification in his sense to be a necesssary condition of knowledge, and that he holds truth also to be a necessary condition.(27) Then knowledge turns out to be (at least) true belief that is generated by adaptive explanatory techniques. But this seems odd; now that we recognize two primitive epistemic values, they may well conflict. For example, is it good to believe, in some circumstances, highly explanatory falsehoods? The account surely leaves some such cases strictly undecidable, since it describes both elegance and truth as intrinsic values. But is this plausible? Surely, we might want to say, though it can be useful to believe all sorts of falsehoods, it is always epistemically good to believe the truth. It may be useful, for example, for me to have a cognitive technique that causes me to believe that I have all sorts of positive qualities to an extremely high degree; I may be happier and more efficient if I believe myself to be extraordinarily intelligent, good-looking, humble, and so forth. But it is not a good thing epistemically to believe such things if they are false. (Luckily, however, in my case they are all true.) In short, we seem to have a notion of epistemic value which is radically divorced from utility.


This raises the fundamental difficulty with Lycan's strategy; it makes prudential justification a component of knowledge. But if, as I say, we have a notion of epistemic justification in distinction from prudential justification, and if knowledge is essentially an epistemic matter, then Lycan's account leaves us confused. Either we ought to believe what is produced by procedures that are adaptive or else we ought to believe what is produced by procedures that are likely to yield true beliefs. But we cannot coherently demand that we follow both of these as ultimate aims, because they may and in fact will conflict. Then we are left with an internally incoherent concept of knowledge. Of course, we can subsume one to the other; we can assert that the truth is adaptive and that is why it should be believed, or that we should believe what works on explanatory grounds because it is likely to be true. But then, first of all, we need an account of why this should be so. And second, we should now characterize knowledge univocally as belief in what is true, or belief in what is adaptive. If we take the second road, then I assert that we are going to hold all sorts of falsehoods to be known, and that is prima facie implausible.


In fact, Lycan keeps coming back to the suggestion that explanatory virtues are truth-conducive, although he often enough admits that there is no reason why this should be the case. This is because, I think, he shares the deep-seated intuition that knowledge and truth are bound up, that knowledge is a distinctively epistemic notion. He writes:



Nor, please note, do I understand "rational" here as "likely to yield truth," for this would be . . . to introduce "true" at the beginning of our epistemological quest rather than see it fall out at at the end. . . . [I]ntroducing "true" at this point would both threaten circularity or regress again and invite questions of the form "But what reasons have you to think that U leads to truth?" where U is a posited ultimate principle of theory choice. . . . In the beginning, "rational" is a primitive term used to evaluate epistemic acts; particular principles are later seen to "tend toward truth," because the beliefs they produce are rational - not the other way around. (155,6)







This passage strikes me as wanting to have it both ways. It seems that we simply ought to believe what is rational (simple, powerful etc.), and then, as a sort of added bonus, as a matter of fact what we believe will happen to be true. But this is odd if, as Lycan admits, there is no reason to think that a simple explanation is more likely to be true than a very complex one. Now as a matter of fact, the notion of truth itself here may well be disintegrating; Lycan's pragmatist tendencies might be extending toward taking truth to be warranted assertibility. The view would then face familiar and, I think, devastating problems, but I will not address those here.(28)


But there is a further argument that Lycan gives for his position. He argues that we cannot be enjoined from doing what we cannot but do, and that we cannot fail to use explanatory criteria in generating beliefs:



I suspect that it is biologically necessary to a sentient organism's survival that the organism organize its experience in the simplest, most coherent and expedient, and yet most powerful and comprehensive way that it can. . . . It thus comes to be be lawlike that a viable organism operates, epistemically, according to the elegance principles . . . or something like them. . . . If the suspicion I have expressed is well-founded, the point is this: A normal human being cannot keep from forming beliefs in ways that have survival value, or from making use of the elegance principles in doing so - at least up to a point. And if that is so, then no one is obligated in any way to refrain from following those primitive procedures. (126, 7)







I would like to make several remarks about this passage. First of all, there is no reason whatever to believe that it is true that we are under adaptive necessity of adopting the most elegant explanation we can come up with, which is exactly what Lycan implies. In fact, that is a bizarre assertion; it claims quite generally that people never, or at any rate very rarely, adopt a less elegant hypothesis when a more elegant one is available, which is obviously false. Aside from the questions about the nature of simplicity (which Lycan leaves more or less completely obscure, and which I suspect cannot be clarified), I can surely resolve to adopt whatever hypothesis seems most likely to be true, however inelegant it may seem to be. It might be very elegant indeed, for example, to believe that the extinction of the dinosaurs resulted from a nuclear test on this planet by aliens, but a nice messy hypothesis involving climactic changes and disease might seem more likely to be true.


In fact, Lycan here goes further in his doxastic determinism than does even a rabid determinist such as Alston. For Alston admits that we can do our best to apply conscientious techniques of belief-formation. He simply denies that the fact that this is to some extent under our control is enough to render the beliefs that such techniques generate voluntary and hence subject to praise and blame. But Lycan asserts that we simply must use the techniques of belief-formation that we do use. That is false. Advances in methodology have been made, and can be voluntarily adopted. Nor do they all yield increases in elegance.



XIII. The Epistemic Telos


I think the case against Lycan's view, and in fact the case against any view which separates truth from justification can be neatly clinched if, as I have argued elaborately in the first chapter, belief is always verifically oriented. Again, knowledge is our goal in the formulation of particular propositional beliefs. However, to believe a proposition is (with appropriate refinements) simply to take that proposition to be true. This shows that, in fact, our telos with regard to belief is simply to believe the truth. This is necessarily the case in every instance of belief. And if it is the case, and if knowledge is the epistemic telos, then knowledge is merely true belief.


Now it is obvious that believing some p may in fact conduce to various goals of mine aside from truth. For example, believing that p may make me happy, or it may be adaptively useful, or it may keep me from experiencing a psychological breakdown. And all of these goals may in fact operate as motivations for me to believe p; they may be causes of my belief that p. But such goals operate in a fundamentally different way in my doxastic procedures than does the goal of believing the truth. To see this, note that, even if see very clearly that p is adaptively useful, I will not believe it until I take it to be true. Now there are various ways in which I can begin to make myself take p to be true in such a situation. But every one of them is a way in which I can come to take p to be true. That is, every one of them is an instance of what, in the first chapter, I called self-deception. That is, I can only believe p when I come to take p to be true, when the telos of the propositional attitude is experienced as truth.


If I do not, for example, think that elegance, simplicity, and so forth are indications of truth, then the observation that some hypothesis is elegant and simple brings me no closer at all to believing it than I was already. John Stuart Mill, in arguing that happiness is the telos of human life (a claim which, by the way, is false if, as I have implied, true belief is an intrinsic goal which is divorced from utility), said this:



The only proof capable of being given that an object is visible is that people actually see it. The only proof that a sound is audible is that people hear it; and so of the other sources of our experience. In like matter, I apprehend, the sole evidence it is possible to produce that anything is desirable is that people do actually desire it.(29)



I take this to be true. And the only proof that the goal of inquiry is true belief is that this goal is actually in view in every case of belief-generation. I think that can be shown to be the case. Again, I will stake the matter on this: if you can believe any p which you do not take to be true, but which you take to be adaptively productive, conducive to happiness, hilariously funny, etc., then you will have refuted the current position.



XIV. A Dilemma


Finally, I would like to formulate the argument that knowledge is merely true belief as a dilemma. If accounts of justification themselves must be meta-justified by an argument to the effect that beliefs justified on the account are likely to be true, then justification is of instrumental rather than intrinsic value. That is, we want justification with regard to particular propositions not because of its intrinsic worth, but because it is a means to what we regard as having intrinsic worth, namely true belief. If knowledge is the epistemic telos with regard to particular propositions, then it follows from such a position that knowledge is merely true belief. On the other hand, if justification is valued not for its truth-conduciveness, but for its conduciveness to some other goal, for example, successful adaptation, or for that matter, if justification is itself proposed as an intrinsic goal (a demand of reason, for example), then knowledge is an incoherent notion. It gives us two goals for inquiry, which cannot always be realized simultaneously. As far as I can see, these two possibilities exhaust the possible accounts of knowledge as (at least) justified true belief. Again, either justification is instrumental to truth or it is not. If it is, then knowledge is merely true belief. If it is not, there is no longer a coherent concept of knowledge. Thus knowledge is merely true belief. Q.E.D.





In chapter two, I used Moore, Diogenes, Johnson, Kierkegaard, Nietzsche, James, and Santayana to argue that the claim that justification is logically necessary for knowledge is false to our commitments and practices as doxastic agents. This emphasis on doxastic agency followed from the characterization I gave in chapter one of belief and doubt as relations of particular agents to propositions. And indeed, to repeat, I do not know of any philosopher who does not hold that every belief is someone's belief. Now the ultimate reason why I think that knowledge is merely true belief is simply this: it is far truer than the epistemology of justification to our lives as doxastic agents. In my view, all of the doxastic agents we know about are organisms scurrying around on the surface of a planet. We have the commitments necessary to that situation; the world constrains us for the most part to the truth. There is no reason to think that all knowledge is rationally justifiable, especially if, as I argued in chapter three, our relations to the world are fundamentally non-epistemic: not a matter of inference (e.g.) from an internal component of experience. The doxastic agents with whom we are acquainted are full-fledged organisms, not the bits of those organisms that reason. Thus, I construe that thesis that knowledge is merely true belief as a simple acknowledgement of reality.


1. However, for a vaguely similar strategy, see Joseph Margolis, "Alternative Strategies for the Analysis of Knowledge,"


Canadian Journal of Philosophy 2 (1973): 461-469. Some philosophers have denied that justification is a necessary condition for knowledge. See, e.g., Peter Unger, "Experience and Factual Knowledge," Journal of Philosophy, vol. 64 (1967), pp. 152-173. William Alston, "Justification and Knowledge," Epistemic Justification (Ithaca: Cornall University Press, 1989), pp.172-182. Fred Dretske, Knowledge and the Flow of Information (Cambridge, Mass.: MIT Press, 1981), chapter 4. But these philosophers do add a third condition to the analysis of knowledge. In fact, I would say roughly that they all regard justification as internal, while the condition they adduce for knowledge beyond truth and belief is external. Of course, other philosophers simply take such external conditions to be accounts of justification. In what follows, such views are implicitly included among the traditional analyses.


2. "The term `justification' in its pre-analytic sense may be thought of as being the name for that which distinguishes knowledge from true belief which is not knowledge. . . . We presuppose, then, that there is a valid distinction between knowledge and true belief which is not knowledge." Roderick Chisholm, "A Version of Foundationalism," Midwest Studies in Philosophy 5: 1980, p.560.


3. As we saw in chapter one, according to R.B. Braithwaite `I believe that p' means "(1) I entertain p" and "(2) I have a disposition to act as if p were true." "The Nature of Believing," Proceedings of the Aristotelian Society, vol.33 (1932-3), p.132.


4. This has been pointed out by, among others, H.H. Price. "Some Considerations About Belief," Proceedings of the Aristotelian Society, vol.35 (1934-35), pp.236 ff.


5. See, e.g., Laurence BonJour, The Structure of Empirical Knowledge (Cambridge Mass.: Harvard University Press, 1985), p.4. Paul Moser, Empirical Justification (Dordrecht: D. Reidel, 1985), p.22.


6. "The theory of knowledge is an attempt to answer the question `How do you know?'." . . . In asking how a person knows something we are typically asking for his grounds for believing it. We want to know what justifies him in holding his belief." John Pollock, Contemporary Theories of Knowledge (Totowa, New Jersey: Rowman and Littlefield, 1986), p.7.


7. A mathematician of my acquaintance describes her process for solving problems in just this way. She reports that she knows the solution to the problem well before she can formulate a proof, and often before she has any story at all to tell about why the solution is right. This is by no means atypical. See Henri Poincare, "Mathematical Creation," trans. G. Bruce Halstead, The Foundations of Science (New York: The Science Press, 1913). Jaques Hadamard, The Psychology of Invention in the Mathematical Field (Princeton: Princeton University Press, 1945).


8. See, for example, Bonjour, The Structure of Empirical Knowledge, chapter 1. Alvin Goldman, Epistemology and Cognition (Cambridge, Mass.: Harvard University Press, 1986), pp. 116-121. Moser, Empirical Justification, pp. 4-8.


9. David Armstrong, Belief, Truth, and Knowledge (New York: Cambridge University Press, 1973), p.173.


10. This is a simplistic version of the view advocated, e.g., by Goldman in Epistemology and Cognition.


11. Armstrong, Belief, Truth, and Knowledge, chapter 12.


12. For example, Chisholm, "A Version of Foundationalism."


13. Bonjour, The Structure of Empirical Knowledge, chapters 5, 6, 7.


14. Here, I pre-suppose that a solution to a problem must be true. Goldman has defended this position in Epistemology and Cognition, pp. 125-131.


15. This objection was brought to my attention by Tim McGrew.


16. On internalist views that agent must be aware that the belief was so produced; on externalist views the agent need not be aware of this fact.


17. In fact, I agree with Alvin Goldman that they are very intimately connected indeed. See Epistemology and Cognition (Cambridge Mass.: Harvard University Press, 1986).


18. My characterizations of these ethical positions is sketchy and inadequate to many actual positions. The present point, however, is simply the parallel to epistemology, so we need not be concerned with capturing the subtleties of full-fledged ethical theories.


19. William Alston, "Concepts of Epistemic Justification," Epistemic Justification (Ithaca: Cornell University Press, 1989), pp.81-114.


20. Alston, "Concepts of Epistemic Justification," Epistemic Justification, p.96.


21. See especially "The Deontological Conception of Epistemic Justification," Epistemic Justification, sections II-VI.


22. Paul Moser, Empirical Justification (Dordrecht: D. Reidel, 1985), p.4.


23. Goldman, Epistemology and Cognition, pp.116-121.


24. Laurence Bonjour, The Structure of Empirical Knowledge (Cambridge, Mass.: Harvard University Press, 1985), p.8.


25. Goldman, p.106.


26. William Lycan, Judgement and Justification (Cambridge: Cambridge University Press, 1988), p.148.


27. It is only on this assumption that Lycan's position constitutes an alternative to the views described in the previous section. So I shall talk as though this is Lycan's position, though, again, he does not explicitly endorse it.


28. All that would be necessary to refute this view is the description of a single case in which some agent is warranted in asserting p and p is false. Such cases are not hard to devise. Lycan assures me that this is not his view of truth.


29. John Stuart Mill, Utilitarianism (Indianapolis: Bobbs-Merill, 1985 (1861)), p.44.