In this essay I shall explore the possibilities for knowledge of God that are opened up by recent developments in epistemology that go under the title externalism; more specifically, I shall be concerned with the version of externalism known as reliabilism. I shall set this up with a consideration of how those possibilities look from a more internalist epistemological stance. I shall be working from within the Christian tradition, though I take my remarks to have a wider bearing.

It is a familiar view that knowledge of God — His nature, doings, and purposes — is either nonexistent or very restricted, and that, at least for the most part, believers have to make do with faith rather than with knowl edge. This view has been widely held by both friend and foe of religious belief, and it goes back many centuries. A classic statement is found in Aquinas. Although the existence and certain basic features of the nature of God can be known in the strict sense, everything else in the Christian faith must be accepted on faith, which he defines as "the assent of the intellect to that which is believed" where this is "not through being sufficiently moved to this assent by its proper object [in which case it would constitute knowledge], but through an act of choice, whereby it turns voluntarily to one side rather than to the other." Faith is distinguished from opinion by the fact that it involves "certainty and no fear of the other side."

In modern times the view has been enthusiastically endorsed by such diverse fideists as Pascal, Kant, and Kierkegaard, but its acceptance has been much wider, including, on the side of believers, such figures as F. R. Tennant and Paul Tillich. Needless to say, unbelievers deny any knowledge of God; we will attend to some of them shortly. Deliberately distorting the terms for my own use, I shall refer to those who hold that we can have knowledge of God as gnostics and those who deny this as skeptics. I shall set aside questions about the meaningfulness, consistency, coherence, and other internal features of religious belief, taking it that all parties to the discussion agree that beliefs about God constitute genuine candidates for knowledge, if only the right further conditions are satisfied.

Whether we can have knowledge of God depends on what knowledge is, and varying positions on this have been taken in the history of philosophy. Throughout most of that history, the dominant conception has been what we may call the intuitive conception. Knowledge, to quote a particularly succinct twentieth century formulation, is "simply the situation in which some entity or some fact is directly present to consciousness." Various forms of this conception are to be found in Plato’s conception of the awareness of Forms, Descartes’s notion of "clear and distinct perception," and Locke’s definition of knowledge as the "perception of the agreement or disagreement of ideas," as well as Aquinas’s insistence that for any knowledge that goes beyond sense perception, the object must be intellectually "seen," that is, as directly presented to the intellect as a seen object is to vision. Many contemporary epistemologists take this notion to be intolerably obscure; nevertheless, I think that paradigm examples like simple, self-evident truths of mathematics and logic tie down the notion sufficiently to enable us to see that articles of the Chris tian faith like Creation, the dealings of Yahweh with Israel, the Incarnation, and the Trinity are not strong candidates for being directly presented to the mind in this way. To be sure, the propo nents of this conception of knowledge recognize that a fact may be known not only by being itself directly presented to awareness but also by being logically derived from such truths. That leaves open the possibility that some articles of faith might be demonstrated from premises that are directly presented to awareness, but, by almost common consent, this is a live possibility for a few such articles at most, and the usual contemporary view is that it extends to none.

This denial of knowledge of God is only as cogent as the conception of knowledge on which it is based, and that conception has virtually disappeared from the epistemological scene. One factor in this is disillusionment about the concept of direct awareness, presentation, or givenness of facts,⁶ though, as indicated above, I find this reaction to be overblown. It is true that, by the nature of the case, no satisfactory analysis can be given of the notion of direct awareness, but it is only an unfounded dogma that every respectable concept must justify that title by the provision of an analysis; analysis must start somewhere, and why not here? A more serious disability is the drastic restriction on the extent of knowledge that this conception enforces. Very little of what we take ourselves to know can lay claim to be "directly present to consciousness." The past and future, the distant in space, generalizations of all sorts, hypothetical the unobservable fine structure of things, all fall outside this area. Knowledge is restricted to self-evident truths, our own conscious states, perhaps what we directly perceive in the external environment, and what can be deductively derived from this. Thus H. A. Prichard:

…we are forced to allow that we are certain of very much less than we should have said otherwise. Thus, we have to allow that we are not certainof the truth of an inductive generalization, e.g., that all men are mortal, or that sugar is sweet, for we are not certain that anything in the nature of man requires that he shall at some time die; we are not even certain that the sun will rise tomorrow . . . It is of no use to object, "Well, if you are going to restrict what we know to what we are certain of, you are going to reduce what we know to very little." For nothing is gained by trying to make out that we know when we do not….

But we are forced to recognize that we do not know any of these things only if we adopt a conception of knowledge as restrictive as this one, and that is just the question.

II. The Justified-True-Belief Conception and Knowledge of God

In this century, at least in English-speaking circles, the assessment of the possibility of religious knowledge is more likely to be conducted against the background of a conception of knowledge as justified true belief (JTB). Whereas the intuitive conception took knowledge and belief to be mutually exclusive psychological states (belief being the supposition that a fact obtains, as contrasted with seeing that it obtains, having the fact directly presented to one’s awareness),⁸ the JTB view regards knowledge as a belief that passes certain tests, namely, truth and justification. Clearly this requires a different conception of belief as well as of knowledge. The JTB conception works with a more neutral account of belief, one that can be briefly indicated by saying that a sufficient condition for a normal mature human being, S, to believe that p is that S would give a positive response to "Is it the case that p?" provided S were disposed to be candid and cooperative.⁹ This condition will be satisfied whether S knows or "merely believes" that p.

Before exploring the use of a JTB conception of knowledge in the rejection of knowledge of God, we ought to note the general acceptance of Edmund Gettier’s demonstration of the insufficiency of these conditions for knowledge in his celebrated article, "Is Justified True Belief Knowledge?"¹⁰ Contemporary versions of the theory always include one or more additional conditions designed to forestall Gettier counterexamples. I shall ignore this complication in the sequel, since the attacks on religious knowledge in which I am interested are all designed to show that the putative knowledge in question fails to satisfy the justification condition. Let me also say a word as to how we should understand ‘justification’ for purposes of this discussion. (Here I speak not of the conditions for justification, which I will discuss at length, but of what justification is, how we are to conceive that for which the alleged conditions are conditions.) The major divide here is between those who do and those who do not take being justified in believing that p to consist in some sort of "deontological" status, for example, being free of blame for believing that p or having satisfied one’s intellectual obligations in doing so. The nondeontologists will generally take justification to be some other sort of evaluative status, for example, being based on an adequate grounds In the works cited in notes, I give reasons for rejecting any version of a deontological concept, on the grounds that they either make unrealistic assumptions of the voluntary control of belief or they radically fail to provide what we expect of justification. Hence, I will be thinking in terms of some nondeontological evaluative conception.

I may as well go with my favorite: "being based on an adequate ground." As we shall see, it fits nicely into the arguments given by our contemporary skeptics. In looking at twentieth-century skeptics, I am going to concentrate on nonbelievers. The main reason for this is that they are more thoroughgoing in arguing for their skepticism. Believing skeptics are primarily concerned with working out a viable alternative stance toward the articles of faith. They tend to quickly concede that knowledge is impossible and pass on to their main task of depicting what we have instead. The unbeliever, on the other hand, is confronted with no such task, and takes his main job to be demolishing pretensions to knowledge of God. As a background for this consideration, let us fill out a bit more the Thomistic picture of the situation. Certain basic propositions concerning the existence and nature of God can be demonstratively proved on the basis of premises that are known with certainty. For other articles of faith, we can produce supporting considerations that make it rational to accept the thesis that their truth is vouched for by the authority of God, though this support is not of such a magnitude as to compel rational assent. The considerations in question appeal to evidences of the divine authority (divine guidance) of the Bible and the church. They include miracles, prophecies, and the growth of the church.

Our contemporary skeptics attack both parts of the Thomistic scheme. They seek to show that alleged proofs of the existence and nature of God lack cogency, and they argue that the "evidences" of reveIation are much too weak to do the job. Just to pick two names out of a vast crowd, this is the sort of thing we find in such works as God and Philosophy and The Presumption of Atheism by Antony Flew and The Miracle of Theism by J. L. Mackie. To have something relatively specific to work with, let us leave to one side the existence, omnip otence, and omniscience of God, and other standard theses of natural theology, and concentrate on a distinctively Christian thesis like the Incarnation. Let us say that the average Christian believes that God became man in Jesus Christ to save us from sin and death (if you prefer some other statement of the purpose, fill that in as you see fit) because this is asserted in the Bible, by Jesus and others, or because it is proclaimed by the church. Our skeptics will take this to be a radically insufficient reason for the belief — radically insufficient to render the belief justified; and since knowledge requires justified belief, this is enough to prevent the believer from knowing this, even if the belief is true. What basis do they have for this judgment? To spell this out adequately, we will have to go into their background epistemology, and to accomplish that, we will have to do a lot of digging. Unlike Aquinas, the likes of Flew and Mackie do not put their epistemological cards on the table. I will not have time for close textual exegesis. I will have to content myself with plausible conjectures that I am confident could be supported by further scholarship. First, we will have to decide whether they are more coherentist or foundationalist. Here I will just say that they certainly do not sound like coherentists.

In any event, if they are proceeding on a coherentist basis, the game is up already, both because of the fatal disabilities to which coherentism is heir and because of the fact that beliefs about God seem to do as well as anything else on a coherentist epistemology I will assume that they presuppose some kind of foundationalism (the exact brand does not matter) and proceed accordingly.

On a foundationalist epistemology, what does it take for a belief to be justified? Broadly speaking, there are two possibilities. A belief can be indirectly or mediately justified, justified on the basis of (by the mediation of) other justified beliefs; or it can be directly or immediately justified, justified on the basis of something other than that. Experience is one major alternative. What gives foundationalism its distinctive thrust is that (1) it recognizes direct as well as indirect justification, and (2) it holds that all indirect justification traces back eventually to directly justified beliefs. Now remember that we are working with a conception of justification as being based on an adequate ground. Thus, if a belief in the Incarnation is to be justified, it must be either based on adequate reasons, in the shape of other things one knows or justifiably believes (indirect justification), or based on something else, for example, on experience, in the way beliefs about one’s current conscious states are, and, depending on one’s views of the epistemology of perception, in the way one’s belief that one sees a tree directly in front of one is (direct justification). As for the latter, contemporary skeptics typically take it as obvious that no religious beliefs can be directly justified. This view has been recently challenged both by Alvin Plantinga¹⁹ and by me.²⁰ Indeed, it is challenged by the entire mystical tradition. This is not to say that every religious belief can be reasonably thought of as justified by one’s experience of God, but neither can it be taken as obvious that no religious beliefs can be directly justified. However, since I have other fish to fry in this essay, I will not contest the point here. I will concede for the sake of argument that if a belief about God is to be justified, the belief will have to be held on the basis of adequate reasons. We must now consider what is required for that.

JTB theorists are by no means in agreement on this, but I believe that an account of the sort I am going to present is widely accepted, and that it is something like this account that lies behind the skeptics’ claim that such reasons as Christians have for their beliefs are insufficient to do the job. Let us say, then, that S‘s belief that p is based on an adequate reason, q, iff:

1. S believes that q.
2. S is justified in believing that q.²¹
3. S believes that p because he believes that q.²²
4. q provides adequate support for p.²³
5. S knows, or justifiably believes, that q provides adequate support for p
6. S would cite q in justification of his belief that p if challenged.²⁴

Let us apply this account of mediate justification to the question of whether it is possible for someone to be mediately justified in believing that Christ died as an atonement for our sins, where the person, S. believes this on the ground that it is asserted by Saint Paul or that it is a dogma of the church. We can stipulate that conditions 1, 3, and 6 are satisfied.²⁵ That is, S believes as he does because of certain reasons and is quite aware that he does so. He has access to the reasons on which his belief depends and he regards those reasons as sufficient. In such cases the claim of the critic is that conditions 2 and 4 cannot both be satisfied. In this instance, for example, condition 2 may be satisfied if the reason is simply that Saint Paul or the church asserts this. There is no particular difficulty in S’s being justified in supposing that. But then condition 4 (and hence condition 5) is not satisfied. The mere fact that the thesis is propounded by that person or that institution is not a sufficient reason for accepting it. We lack sufficient reasons for supposing that Saint Paul or the church authorities have such expertise in theological matters that their pronouncements are a reliable guide to the truth. To be sure, to go into these matters, we would have to examine the traditional "evidences" of the authority of the apostles and of the church and determine whether they are sufficient to shore up the claims they are invoked to support. But since I am primarily interested in exploring the prospects of knowledge of God on a quite different approach to knowledge, I will just accept the critic’s judgment on these points and agree that no one has sufficient reason for supposing that condition 4 is satisfied in this instance (and hence that no one satisfies condition 5). To be sure, if q were beefed up to include the proposition that the Bible is the word (message) of God to us or that the church is guided by the Holy Spirit in its doctrinal pronouncements, it would be a different story. Now it will be universally agreed that the reason provides adequate support. Surely God (the Holy Spirit) is an authority on theological matters! But now the critic will deny that S is justified in accepting this enriched premise and deny this for essentiall y the same reasons for which he denied that condition 4 is satisfied for the original premise. Just as we lack sufficient reason for supposing that the mere fact that p is asserted by Saint Paul or the church is a sufficient reason for believing it, so we lack sufficient reason for supposing that the Bible is the word of God or that the church is guaranteed by God to be correct in its doctrinal pronouncements. If we had sufficient reason for the latter, we would ipso facto have sufficient reason for the former.²⁶

So let us agree that, by the standards of this epistemology, Christians are never justified in their distinctively Christian beliefs about God, and hence that, even if those beliefs are true, none constitutes knowledge. Whether this should disturb Christians depends on the credentials of this epistemology, a matter to which I now turn.

III. Internalism

The most obvious problem facing a JTB conception of knowledge, and indeed facing any theory of justification, is a methodological one: How do we determine what conditions are sufficient for justification? (Now we are considering what, in particular, is required for justification, not how the concept of justification is to be understood.) In a foundationalist theory, this divides into two questions: (1) How do we tell what suffices to render a belief directly justified? and (2) How do we tell when a reason provides adequate support for a belief (adequate grounds for taking the be lief to be true)? These are very large questions, and here I will be able only to indicate the most important difficulties confronting TB theorists in this area.

One approach to these issues has to do with likelihood of truth. After all, whatever else epistemic justification of beliefs may be, it is clearly supposed to be a commodity that is valuable from the standpoint of the search for truth, for the aim at restricting our beliefs to what is true. This suggests that a belief counts as justified only if it is at least likely to be true and, more specifically, that what justifies a belief thereby renders it likely to be true. And that suggests in turn that at least part of what is required if a condition C is to suffice to justify belief B is that the process of forming B on the basis of C is a reliable one, one that can be counted on to lead to truth, at least most of the time. So if we want to know whether a belief that there is a tree in front of me is justified by virtue of being based on a certain kind of visual experience, what we need to consider is whether the process of belief formation exemplified by forming that belief on the basis of that kind of experience is a reliable one, one that can be depended on to (usually) yield true beliefs. And if we want to know whether the belief that the generator is not functioning properly is justified by virtue of being based on my knowledge that a certain portion of the dashboard has lighted up, we must consider whether beliefs like that are reliably formed on the basis of facts like that; in other words, we must consider whether such a light is a reliable indication of a defective generator.

But the trouble with this, a trouble that, among other things, has inhibited most epistemologists from proceeding in this way to determine what justifies what, is that we cannot settle these questions about reliability without relying on principles of justification that are of the same sort as those we are trying to establish. How do we tell whether a visual experience of type V is a reliable indication of there being a tree in front of me? We certainly have to use empirical evidence to do so, which means that we have to suppose ourselves to be justified in holding a number of beliefs about the environment on the basis of perceptual experience. And how do we tell that a certain light is a reliable indication of a defective generator? Again, by empirical investigation that will involve assuming, in many cases, that we can form a justified belief on the basis of like indications. Hence, we can never get started on such investigations, at least not without circularity. We have to make use of some principles of justification, at least assume them in practice, to establish any of them in this fashion.

Hence, most theorists of justification have shied away from looking to considerations of reliability to determine what justifies what, and they have taken another route. They have supposed that such determinations can be made, at least in the most basic cases, by rational reflection, just by carefully considering the matter, by "armchair thinking." Let us call this approach internalism, using the term in one of the ways it is used in contemporary epistemology.²⁹ Thus Roderick Chisholm:

Now, I think we may characterize the concept of "internal justification" more precisely. If a person S is internally justified in believing a certain thing, then this may be something he can know just by reflecting upon his own state of mind. And if S is thus internally justified in believing a certain thing, can he also know, just by reflecting upon his state of mind, that he is justified in believing that thing? This, too, is possible — once he has acquired the concept of epistemic justification.³⁰

On this "internalism," justification is something of which we can have a priori knowledge, just by thinking about the matter. Like intuitionism in ethics, the view comes in two versions: general and particular. According to the former, we are capable of discerning the truth of general principles of justification by reflection. According to the latter, a more popular form with contemporary internalists including Chisholm, what we can discern on reflection is that a particular belief is or is not justified.³¹ We can then arrive at correct general principles of justification by determining which principles best accommodate particular facts of justification.

Thus, to return to our focal concern with Christian beliefs, a skeptic who is an internalist JTB theorist, closet-variety or otherwise, will suppose that he can determine just by careful reflection that, for example, introspective and perceptual beliefs are justified when formed in the usual way, and that beliefs formed by valid deductive or sound inductive inference from justified premises are thereby justified. But he will report that even on the most careful scrutiny he cannot see that a belief is justified by virtue of being asserted in the Bible or by the church, unless the credibility of these sources has been established on the basis of beliefs that, like those just mentioned, can be seen on reflection to be justified.

IV. Troubles with Internalism

Once again, this denial of knowledge of God is only as compelling as the epistemology on which it is founded; and what are we to say about that epistemology? There are serious reasons for being suspicious of internalist justification theory, and hence of an internalist JTB account of knowledge. The most fundamental weakness of internalism is its indeterminacy and arbitrariness. First, consider the particularistic variety. Just what is the status of the conviction that a particular belief in particular circumstances is or is not justified? Reflecting on the sharp divergencies in such convictions among philosophers, an unsympathetic observer will Suspect that these judgments often reflect the theoretical prepossessions of the one who is judging.

Am I justified in believing that there is a tree in front of me when I take, or would take, myself to suppose that I see a tree in front of me, in the absence of any reasons for this? Chisholm would say yes; those who stress the need for discursive support for perseptual beliefs — C. I. Lewis, Wilfrid Sellars, Laurence Bonjour — would say no. Does the mere fact that I believe something that is not contradicted by the sum total of my other beliefs render me somewhat justified in believing it? Some will reply in the affirmative and others in the negative And so it goes. This is too easy a way out, too easy a way to put the stamp of rational acceptance on one’s predilections. Where this approach does seem to yield definite results, this only reflects the fact that most of us are predisposed to agree about the conditions under which certain kinds of beliefs are to be approved. And intuitionism with respect to general principles of justification seems even less promising. Can I ascertain just by reflection the conditions under which perceptual judgments are justified? Again, a consideration of sharply differing positions on the epistemology of perception may well lead us to skepticism about this. Insofar as mere reflection does seem to yield definite and relatively uncontroversial results, it yields far too little to cover knowledge that it seems for all the world as if we possess. Perhaps we can determine just by thinking about the matter that beliefs about one’s current conscious states and beliefs in simple self-evident truths of logic and mathematics are ipso facto justified. But it is dubious that the justificatory status of anything else is clearly revealed to an unclouded inner gaze. In fact, those who have taken intuitionism most seriously in epistemology from Descartes on have generally supposed that the most one can see to be justified immediately in the area of sense perception, for example, is that one is currently having such and such sensory experiences; the existence and nature of anything external will have to be established by argument. It is too well known to require mention that the chief burden of Western philosophy from Descartes on has been the task of "proving the existence of the external world."

Those who have sought to evade this task, from Reid to G. E. Moore and H. H. Price, have, while taking (some) perceptual beliefs to be immediately justified, definitely not claimed an intuition of the truth of any such principle of justification. In stead they have defended it as a principle of commonsense³² or a reasonable assumption³³ or the only alternative to skepticism.³⁴ The lack of intuitive assurance for any principles of inductive logic has been, if anything, even more generally recognized since the time of Hume. Thus if we rely on intuitionis m we will be in danger of ending up with a basis of private experience and deductive logic — a basis far too meager to yield anything like the extent of knowledge it seems clear that we possess.

Some intemalists have sought a firmer basis for principles of justification than the "bare intuition" envisaged by such thinkers as Chisholm, while still preserving the a priori character of our knowledge of justification. One suggestion is that what suffices for the justification of beliefs in a certain range is determined by the concepts used in beliefs of that range. The idea goes back to Wittgenstein’s insistence on "criteria" for the application of concepts. A prominent contemporary exponent, who stresses the epistemological bearing of the view, is John Pollock.

To learn the meaning of a concept … is to learn how to use it, which is to learn how to make justifiable assertions involving it…. the meaning of a concept is determined by its justification conditions.³⁵

The justification conditions are themselves constitutive of the meaning of the statement. We can no more prove that the justification conditions of "That is red" are the justification conditions than we can prove on the basis of something deeper about the meaning of "bachelor" that all bachelors are unmarried … the justification conditions of "That is red" or "He is in pain" are constitutive of the meanings of those statements and hence cannot be derived from any deeper features of their meanings. There are no deeper features.³⁶

Thus, we can know, just by reflecting on the concepts involved, what it takes to justify any particular belief. And this holds open the promise of an internalist validation of principles of justification.

There are various difficulties in supposing that all, or almost all, of our concepts are made up of justification conditions, but in this brief discussion I will confine myself to the following point. If justification conditions are to make up even part of my concept of a tree, then obviously I must have the concept of justification. And it seems clear that the least sophisticated humans who have the concept of a tree, like very small children, lack any such concept of justification. A two-year-old child knows what doors and windows, birds and dogs, trees and bushes, adults and children look like, and he can recognize them perceptually; hence, he has concepts of these objects. But it would be rash to suppose that the two-year-old can wield a concept of being epistemically justified in a belief. Of course one does not have to be able to express a concept verbally to have it; but it is highly dubious, at best, that a two-year-old child has a concept of epistemic justification even in a tacit, "practical employment" form. And once we recognize that one can have a usable concept of a dog or a tree without any justification conditions figuring in that concept, we are led to wonder whether the presence of such conditions in more sophisticated concepts, if they are present, has the epistemological consequences Pollock draws. Pollo ck supposes that it is conceptually impossible for us to fail to be justified in believing that there is a tree in front of us when we have the right kind of sensory experience (in the absence of reasons for thinking the situation to be abnormal), and that we can realize this to be the case just by reflecting on our concept of a tree. But once we see that it is always possible to retreat to a more primitive concept that is like the (untechnical) adult concept except for embodying no justification conditions , we must consider whether we should use the more rather than the less sophisticated concept.³⁷ And that, at bottom, is just the question of whether the justification conditions involved, by hypothesis, in the adult concept, are, in fact, required for justification. We have not really answered that question by noting that our adult concepts embody an answer to it. We are still left with the question of whether that is the correct (warranted, reasonable) answer. The appeal to the constitution of concepts has gotten us nowhere.

Finally, if one despairs of an a priori grasp of objective facts of justification but still feels tied to internalism, one may seek to subjectivize justification in one or another way. Maximal subjectivity (whatever one thinks justifies a belief does so) has not been popular, but a more qualified subjectivism has been given powerful expression by Richard Foley in his recent book The Theory of Epistemic Rationality.³⁸ On Foley’s position it is epistemically rational for a person to belie ve that p "just if he has an uncontroversial argument for p, an argument that he would regard as likely to be truth preserving were he to be appropriately reflective, and an argument whose premises he would uncover no good reasons to be suspicious of were he to be appropriately reflective." Contrary to appearances, this formulation does not rule out immediately justified beliefs; a belief is immediately justified when the argument in question has the sole premise that the subject believes that p. Now Foley’s view is certainly sufficiently subjectivist for it to be possible that the justification conditions for a given kind of belief are radically different for you and me, in case there are radical differences in what we would judge about arguments on appropriate reflection. Foley believes that there is considerable convergence in the population, but that is not part of the position. But an account of justification that allows for such individual differences does nothing to support the skeptics’ cause. If they were to adopt that account, they would be powerless in the face of a believer who, on adequate reflection, would take himself to have an uncontroversial argument for the Incarnation. Hence, a position like Foley’s can be of no use for the skeptic.

Thus, an internalist account of justification, of the sort under consideration, faces considerable, even fatal, difficulties of its own, quite apart from any problems that specifically concern the religious sphere; and these difficulties are inherited by an account of knowledge that features intemalist justification as a prominent component. When we take justification to be a matter that is to be ascertained, if at all, a priori, by mere reflection, we find our faculties to be unequal to the job of yielding unequivocal, assured results–much less results in sufficient quantity to cover the territory. And this is the case, whichever of the modes of a priori knowledge we think is appropriate to knowledge of justification: synthetic, with Chisholm and others, or analytic, with Pollock and others.

Next I want to note a way in which this internalism constitutes a limitation on the powers of these attempts to determine the conditions of justification, even if we waive the criticisms already aired. The limitation becomes apparent the moment we ask the question, "Can we use these techniques to validate the procedure of forming beliefs on the basis of reflection (rational intuition, reflection on concepts)?" The project of certifying perceptual, memorial, or inferential justification by rational reflection seems attractive just because it avoids epistemic circularity. But that advantage is conspicuously lacking when it is rational reflection itself we seek to ratify. So, at most, the intemalist approach would enable us to validate all sources of justification save one, at the cost of taking the credentials of this source for granted. Once we look at the matter in this way, we are struck by a certain arbitrariness in the procedure. Why should we take the justificatory efficacy of reflection for granted, while insisting that the credentials of other sources must be ratified by the use of the former? Is there any justification for this partiality? Thomas Reid saw this point with stunning clarity and gave it powerful expression.

The author of the "Treatise of Human Nature" appears to me to be but a half-skeptic. He hath not followed his principles so far as they lead him; but, after having, with unparalleled intrepidity and success, combated vulgar prejudices, when he had but one blow to strike, his courage fails him, he fairly lays down his arms, and yields himself a captive to the most common of all vulgar prejudices — I mean the belief of the existence of his own impressions and ideas.

I beg, therefore, to have the honour of making an addition to the skeptical system, without which I conceive it cannot hang together. I affirm, that the belief of the existence of impressions and ideas, is as little supported by reason, as that of the existence of minds and bodies. No man ever did or could offer any reason for this belief. Descartes took it for granted, that he thought, and had sensations and ideas; so have all his followers done. Even the hero of skepticism hath yielded this point, I crave leave to say, weakly, and imprudently…. Chat is there in impressions and ideas so formidable, that this all-conquering philosophy, after triumphing over every other existence, should pay homage to them? Besides, the Concession is dangerous: for belief is of such a nature, that, if you leave any root, it will spread; and you may more easily pull it up altogether, than say, Hitherto shalt thou go and no further: the existence of impressions and ideas I give up to thee; but see thou pretend to nothing more. A thorough and consistent skeptic will never, therefore, yield this point.

To such a skeptic I have nothing to say; but of the semi-skeptic, I should beg to know, why they believe the existence of their impressions and ideas. The true reason I take to be, because they cannot help it; and the same reason will lead them to believe many other things.⁴⁰

The skeptic asks me, Why do you believe the existence of the external object which you perceive? This belief, sir, is none of my manufacture; it came from the mint of Nature; it bears her image and superscription; and, if it is not right the fault is not mine: I even took it upon trust, and without suspicion. Reason, says the skeptic, is the only judge of truth, and you ought to throw off every opinion and every belief that is not grounded on reason. Why, sir, should I believe the faculty of reason more than that of perception? — they came both out of the same shop, and were made by the same artist; and if he puts one piece of false ware into my hands, what should hinder him from putting another?⁴¹

Even though the context of Reid’s discussion is different from ours in that he is reacting to Humean skepticism, the "undue partiality point he is making so sharply is the one I am concerned to make here. Just as Hume and Descartes were exhibiting arbitrary partiality in taking knowledge of one’s own impressions and ideas, plus deductive reasoning, for granted, while requiring that all other claims to knowledge be validated on the basis of the idea is embodied in the contemporary epistemological movement known as externalism. Externalism is the denial of internalism; it essentially involves the lifting of the internalist constraint.

The extemalist holds that it is possible for something to convert true belief into knowledge without the subject’s being aware on reflection, or even being capable of becoming aware on reflection, that it possesses this epistemic efficacy. To be sure, we formulated internalism as a constraint on justification, and so a direct denial would be the thesis that conditions could serve to justify without the subject’s being able to recognize that on reflection. And some externalists do put their position in this way. But others are not concerned with justification. They make a more radical break with internalist JTB theory and leave justification out of the analysis of knowledge altogether. They provide an externalist substitute for justification in the analysis; the most popular substitute is reliability. But even if justification is left out of the account of knowledge, the externalist can still be said to have lifted the internalist constraint on what is required (along with handling Gettier problems) to convert true belief into knowledge, however that is otherwise construed.

In this essay I will focus on reliabilism, construed as the view that knowledge is true reliably engendered belief — again, with some additional condition(s) to take care of Gettier problems.⁴⁵ Justification will be ignored altogether, as far as the analysis of knowledge is concerned; though I have no tendency to deny that most, or perhaps even all, of our knowledge does involve justified belief. As an initial purchase on the relevant idea of reliability, I will just repeat an earlier statement. To say that a belief is reliably produced is to say that it is produced in a way that can generally be relied on to produce true beliefs.⁴⁷ I will develop this position in my own way, but first I will make a few general points that apply to all forms of externalism.⁴⁸

1. Externalism is strongly committed to truth conducivity, the idea that what makes true belief into knowledge is something that will (or would) generally produce true rather than false beliefs. To be sure, the epistemic desirability of truth conducivity is generally recognized, but what we discovered in the previous section was that internalism sacrifices truth conductivity to intuitive recognizability where they conflict, as is often the case, , externalism is prepared to sacrifice intuitive access to truth conducivity It is a question of priorities. On any version of externalism, a belief will not count as knowledge unless there is something about the way in which it is engendered that renders it at least highly likely to be true.

2. As the last remark suggests, externalists hold that how a belief is produced is crucial for its epistemic status. Actually this is an issue on which internalists are divided, some taking the source relevant position just outlined, others holding that a belief is justified provided the believer has sufficient supporting grounds, whether or not the belief is formed on the basis of adequate grounds. But externalists are resolutely committed to the epistemic importance of source.

3. Much of the history of epistemology has been shaped by attempts to respond to skepticism, to show, in the face of skeptical challenges, that we do have genuine knowledge. This may be largely responsible for the pervasiveness of the internalist approach. It has been widely supposed that if I can ascertain in a particular instance whether I really do know just by reflection (intuition), skepticism can get no foothold. This idea may be challenged, as it was by Descartes when he took it that he had to invoke the omnipotence and goodness of God to lay the spectre of skepticism with respect to "clear and distinct perceptions; but the idea has undoubtedly been powerfully influential. Externalists, on the contrary, show little interest in combatting skepticism they typically take the attitude that knowledge (justified belief) is a subject matter that is there to be studied like any other, and that there is no more need to prove the existence of knowledge before developing a theory thereof than there is to prove the existence of plants before developing botany. This s one respect in which externalism is sometimes considered a ‘naturalistic approach to epistemology.

4. Connected with the previous point is the fact that internalist and externalist thinking is shaped by radically different models. The background picture for the internalist is that of the highly sophisticated, reflective person who is concerned to subject his beliefs to critical scrutiny, to determine which of them pass the relevant tests for being justified or for counting as knowledge. Roughly speaking, a belief can not count as justified (as knowledge) unless such a person could be satisfied that it does — hence the intemalist demand for accessibility of epistemic statuses to reflection. The background picture for the externalist, by contrast, is that of theunsophisticated human or animal perceiver, or even a nonsentient recording device like a thermometer, that is receiving or registering information in a reliable fashion, whether or not it is aware that it is doing so, and whether or not it is even capable of raising the question of whether it is doing so. Assuming that such unsophisticated subjects do acquire knowledge, the externalist would seem to be on strong ground in shaping a conception of knowledge that will accommodate them as well as their more sophisticated brethren.

5. For externalism, knowledge that one knows something is much more dependent on other knowledge than is the case for internalism. On the latter position, much of what is needed to determine that a belief is a case of knowledge can be acquired without dependence on other things we know, since we can tell, just by reflection, when a belief is justified. This point can not be pushed all the way, for truth is also required, and for most candidates for knowledge, no one would suppose that the truth value of the belief can be directly intuited. This is the case for all beliefs about the physical world and other persons. However, for externalism not only truth but also what takes the place of justification cannot be determined without relying on other things the subject has learned about the world. On a reliabilist position, the belief counts as knowledge only if the way in which it was acquired is one that is generally reliable, and that is something that one cannot determine just by turning one’s attention to the matter. If it is a perceptual belief, it is a question of how reliable are the perceptual belief-forming mechanisms involved, and that is a question, broadly speaking, for psychology, physiology, and physics. If it is a belief formed by inference, there is, in addition to issues about the way the premises were acquired, the question of whether the principles of inference involved are such as to yield mostly true conclusions from true premises; and, at least for principles of inductive inference, this cannot be determined by armchair thinking. Roughly speaking, what it takes, on reliabilism, to have knowledge in a given area, and whether a particular belief has what it takes, must be determined in the light of what we know about that area. This point is closely connected with (3). Since I can not know whether and when I have reliable belief about the physical world without relying on knowledge of the physical world to do so, it is obviously hopeless to try to meet the skeptical challenge by constructing a noncircular demonstration of the existence of such knowledge. Thus, externalism is quite prepared to countenance epistemic circularity in attempts to show that we have knowledge or to determine the conditions under which we have knowledge.

Now I want to point out the ways in which externalism in general, and reliabilism in particular, is free of the disabilities that we have seen to plague internalism. Most obviously, reliabilism is not lacking in implications for truth; as pointed out above, it gives that consideration the highest priority. If a belief passes the reliable true belief (RTB) test, it cannot fail to be at least highly likely to be true. Second, it is not subject to the indeterminacy or arbitrariness that plagues appeals to intuition. It may, of course, be difficult or even impossible in particular cases to determine the degree of reliability of a certain belief -forming mechanism; but there is nothing arbitrary about the procedure to be employed. We simply make use of whatever resources we have for ascertaining facts in the sphere in question; there is nothing here that affords the same latitude, and temptation, to read one’s theoretical prepossessions into the method.

However, the superiority I most want to stress concems the extent of knowledge that can be recognized on each approach. We saw that internalist theories, to the extent that they do have definite implications, tend to wind up with intolerably narrow limits, with knowledge being restricted to self-evident truths, one’s own conscious states, and what can be deduced from that. RTB theory, on the other hand, can recognize as knowledge whatever true beliefs are formed in a reliable way. There is no a priori guarantee of how much this takes in; but, by the same token, there are no a priori limits. It is to be decided by actual investigation in each putative area of knowledge just what reliable mechanisms are available and what knowledge they do or can produce. Thus, the field is open for perceptual knowledge about the physical environment, knowledge of the past through memory, knowledge of lawlike regularities, and so on.⁵³

In addition to the ways in which externalism escapes the crippling disabilities of internalism, there are other reasons for regarding it favorably. You can think of these as reasons for reliabilism specifically; some of them will also hold for externalism generally, though I will not take time to spell that out.

A. We can think of reliabilism as generated by the marriage of source relevance and truth conducivity. I have already made some remarks by way of recommending truth condusivity. Source relevance can be supported by reflecting on the fact that if I were to possess strong reasons for a proposition but were to accept it on some disreputable basis, I could hardly be said to know it. Suppose that I actually have strong reasons for supposing that Jim is trying to get me fired, and I believe that he is, but not on the basis of those reasons (I am not really activating them from long-term memory), but just because of my paranoia. (I am given to believing this sort of thing about people even without any strong reasons for the belief.) In that case, I could hardly be said to know that Jim is trying to get me fired. Reflection on cases like this strongly supports the thesis that how a belief is acquired, that is, what it is based on, is crucial for its epistemic status.

B. It is plausible to suppose that what makes the difference between mere true belief and knowledge is whether it is just an accident that the belief is true. If I guess who will win the election and just happen to get it right, I can not be said to have known who would win, for it was a lucky accident that I was correct. A true belief that passes the RTB test is clearly not true by accident. If a belief was generated in a way that can be generally relied on to yield true beliefs, that is paradigmatically a nonaccidental way of getting something right. Epistemologists commonly recognize a variety of factors that convert, or contribute to converting, true belief into knowledge. JTB theorists typically think of these factors as justifiers, but to achieve wider coverage, let us think of them as epistemizers, factors that make for knowledge. To focus the discussion, let us concentrate on immediate epistemizers. Popular candidates have been dire ct experience of the object of the belief, the mere truth of the belief for certain kinds of beliefs, the mere existence of the belief for certain kinds of belief, and the self-evidence of a proposition.⁵⁴ But what is it that makes these candidates plausible ones and leaves many other contenders (e.g., being vividly imagined or conforming to one’s desires) outside the pale? An internalist will, of course, recommend a given choice by claiming that its justificatory or epistemic efficacy can be known by intuition or by reflection on one’s concepts. But we have seen that these claims are, at best, terribly inflated; moreover, at best, the alleged reflection does not give us adequate insight into why it is that what justifies does so and what does not justify does not.⁵⁵ Here externalism displays its superiority by providing a simple, unified, and convincing answer to the question of what enables an epistemizer to epistemize: it is the fact that the practice of forming beliefs on the basis of the putative epistemizer is a reliable one. Surely our conviction that a belief is epistemized by one’s being directly aware of what the belief is about is closely tied to our conviction that beliefs so formed are at least highly likely to be true. And the same is to be said for our conviction that any belief to the effect that one is currently in a certain conscious state is ipso facto epistemized; if one did not suppose that such beliefs are either infallible or at least very rarely mistaken, one would not be disposed to take them as epistemized just by being the kinds of beliefs they are. Before turning to the application of externalism to the question of the knowledge of God, I wig fill out the above characterization of a reliable mode of belief formation. This is all the more necessary: since the viability of the notion has been widely questioned of late, and most of these attacks fail to construe the notion in its strongest form.

To say that a particular belief-forming mechanism is a reliable one is not to say that it has a favorable track record. For one thing, it may never have been used and have no track record al all, and yet it may be highly reliable in that it would compile are impressive track record if given a chance. Similarly, a thermometer might be a highly reliable instrument even if it is never used to record temperature. What the reliability of a belief-forming mechanism does amount to is the fact that it would yield a high proportion of true beliefs in a suitable run of employments. What makes a series of uses suitable? For one thing, they must be sufficiently numerous and varied. But they must also involve the kinds of situations we typically encounter. The fact that our normal, visual belief -forming mechanisms would not mostly yield truths in highly unusual environments involving laser images, clever deceptions, Cartesian demons, and the like, has no tendency to show that they are not reliable in the way relevant for epistemic evaluation. Reliability in the kinds of situations that are normal for us is what is crucial. What proportion of truths in a suitable range of cases is needed for a degree of reliability that is sufficient for knowledge? I do not think that there is any precise answer to this question, and insofar as an answer can be given, it may well differ for different areas of knowledge. For purposes of this essay we can think of reliability as involving a high proportion of truths, say, 80 percent or so.

As is implicit in the discussion up to now, it is the reliability of particular belief-forming mechanisms, not the person, generally, that is relevant to the requirements for knowledge. Obviously, a particular person can be very good at forming some kinds of beliefs and very bad at forming others. We do not want to make it impossible for a person to have perceptual knowledge just because he is very bad at doing arithmetical calculations.

A great deal of heavy weather has been made lately over how to individuate belief -forming mechanisms and over how to determine which mechanisms are relevant to the assessment of a particular belief. With respect to my present belief that a computer screen is in front of me, what class of beliefs has to contain mostly true beliefs for this belief to pass the test: visual beliefs, visual beliefs about objects directly in front of one, beliefs about computers, beliefs formed on a Friday, and so on?⁵⁶ Puzzles such as these arise because one ignores the fact that it is belief-forming mechanisms that are in question here, not any old classes of beliefs or belief-forming processes. To know what to look at to ascertain the reliability of a particular belief, we need to determine what input-output mechanism was activated to produce that belief. Otherwise stated, we need to know what input-output function it is by virtue of which that belief was formed on the basis of the relevant input. Still otherwise stated, we need to determine what it was that the psyche was taking account of in forming that particular belief. In this example, the fact that the process took place on a Friday is highly unlikely to play any role in the production of the belief. In fact, it is highly likely that the relevant input consisted of a certain pattern of sensory data, perhaps together with certain background beliefs; anything else true of the current situation was irrelevant to the formation of that particular belief. So the question of whether this belief was reliably formed is the question of whether the activation of a function going from inputs of just that sort to a belief output of just that sort would, in a suitable spread of cases, yield mostly true beliefs.

[To be continued…]