*Originally published in The International Journal for Philosophy of Religion, February 1995

One of the issues that continues to be explored in contemporary analytic philosophy of religion is to what extent (if at all) the claims of Reformed epistemology can be harmonized with evidential requirements for rational or justified belief in God. More generally, what role is there for reasons if theistic belief can be a properly basic belief? In the present paper I will be considering the prospects for the compatibility of Reformed epistemology and evidentialism based on the epistemology of William Alston. After considering two versions of Reformed epistemology, I will argue that both versions are compatible with Alston's multi-level foundationalism and also that his foundationalist epistemology entails a form of evidentialism--higher-level evidentialism. This form of evidentialism imposes an evidentialist requirement for the justification of any higher-level belief about the epistemic status of a putative belief that p. The argument aims to show that imposing such an evidentialist requirement on the justification of higher-level theistic beliefs is compatible with the central claims of RE and satisfies an important desideratum in the evidentialist tradition, the notion of reflective rationality. I conclude that the prospects for the compatibility of Reformed epistemology and an epistemically adequate form of evidentialism are good indeed.

1. Unpacking the Claims of Reformed Epistemology

Over against the so-called evidentialist requirement for theistic belief:

the central claim of Reformed epistemology (hereafter RE) is the proper basicality of theistic belief:

Given the pluriform nature of rationality, the rationality of a basic theistic belief will amount to different things depending on the concept of rationality one is employing. Plantinga and Wolterstorff have construed rationality in [R] as deontological in nature, so that having a rational basic theistic belief will amount to believing in God in such a (basic) way so as to not be in violation of any intellectual duties or obligations.² Alternatively, there is a (a stronger) sense of rationality that requires that a belief be formed in such a way that it is likely to be true, what we might call truth-conducive rationality. We can think of this in terms of either a pure reliability theory (where the reliability of belief-formation is sufficient) or where there is an imposition of a mere reliability constraint (other factors, internalist or externalist, would also be required). This latter option is exemplified in William Alston's concept of epistemic justification and in Plantinga's warrant thesis.³ There are doubtless other ways to think of rationality, though for the purposes of this paper I want to think of rationality (in a strong sense) as at least subject to a reliability constraint. Having said that, it is clear that the concept of rationality I will be working with encompasses Alstonian justification and Plantingian warrant. I will use the neutral word epistemization to range over this reliabilist-rationality.

I do not intend to enter into the issue of what modes of immediate justification are plausible for theistic belief.⁴ Where one requires the belief to be based on adequate (cognitively accessible) grounds, it will be plausible to take the specific mode to be something like being based on the experiential awareness of God (such as in Alston's theory of the non-sensory perceptual experience of God). It could also be construed along purely reliabilist lines and require only the belief's being produced by some reliable mode of noninferential belief-formation. A third possibility, making use of the reliabilist constraint (though within a purely externalist framework), is that we are designed by God in such a way as to naturally form beliefs about Him in certain widely realized experiential circumstances, such as the sight of starry night sky, the majesty of the mountains, or the beauty of a tiny flower, and so on (as Plantinga's warrant thesis maintains). In all of these cases theistic belief is justified or warranted by virtue of nonpropositional evidence, by virtue an evidential relation to something other than some other justified or warranted belief(s) of the subject.

I take it, though, that the claim of Reformed epistemology is actually stronger than [R1]. This can be seen by contrasting it with the notion of proper basicality advocated recently by Anthony Kenny. Kenny has proposed criteria for proper basicality that require there to be evidence for theistic belief. Unlike the traditional evidentialist requirement, his position does not require that the believer possess this evidence, much less base her belief on it. The evidence simply must be available. "So a belief may be basic," writes Kenny, "in the sense of not being held on the basis of reason, but yet defensible to others by the giving of reasons."⁵ This actually forms a premise in Kenny's own argument: "Roughly speaking, a belief can be properly held as basic, without evidence, only if it is rationally defensible. If the existence of God is to be something justifiably held as basic, it must be defensible by argument."⁶ But Kenny's conception of proper basicality is compatible with [R] and [R1], but it is not compatible with what some Reformed epistemologists have had to say about RE's central claim. Plantinga, for instance, says: "Barth joins Calvin and Bavinck in holding that the believer in God is entirely within his rights in believing as he does even if he does not know of any good theistic argument (deductive or inductive), even if in fact no such argument exists" (emphasis mine).⁷ This theme, which runs throughout Plantinga's work, is also emphasized by Stephen Wykstra in his essay "Toward a Sensible Evidentialism" (1989), where Wykstra takes the crux of RE to be, not merely that one's belief in God can be rational without an evidential basis, but that theistic belief can be rational even without there being any evidence available for it.⁸

Hence, a revised version of [R1] gives us:

There are, however, some additional claims made by Plantinga which suggest yet another version of RE. In looking at Calvin, Kuyper, Bavinck, and Barth, Plantinga concludes that "they think the Christian ought not to accept belief in God on the basis of argument" (emphasis mine).¹⁰ Bavinck, Plantinga tells us, emphasizes that we cannot acquire a knowledge of God on the basis of argument, since the theistic proofs do not work. Furthermore, Scripture assumes the existence of God, and so the believer should take belief in God as a starting-point in his reasoning. And thirdly, belief in God is analogous to other beliefs we have (e.g., belief in the existence of the self, the external world, and the past) for which we typically do not have, or need, proof. There is also a sense in which Plantinga sees the Reformed tradition as represented in these four thinkers maintaining that it is inappropriate to believe in God on the basis of propositional evidence, since such evidences cannot produce the certainty which faith requires. Bavinck, for example, held that "the so-called proofs are by no means the final grounds of our most certain conviction that God exists." Plantinga adds that in Calvin's view, "the Christian ought not to believe on the basis of argument; if he does, his faith is likely to be "unstable and wavering," the "subject of perpetual doubt."" Presumably this would make faith "subject to all the wayward whim and fancy of the latest academic fashion."¹¹

So, according to Plantinga, the Reformed objection to natural theology is actually twofold: (a) reasons and arguments are unnecessary for the believer to have a justified belief in God and (b) reasons and arguments are inappropriate as a basis for theistic belief. (A), of course, follows from a rejection of the evidentialist requirement for theistic belief, but (b) is a substantially stronger claim. There is a sense in which (b) is based upon a religious presupposition, which takes the immediacy of belief in (and knowledge of) God as a premise. As Dewey Hoitenga sees it,¹² (b) is a consequence (drawn by the Reformed epistemologist) from the Platonic and Augustinian claim to the immediacy of knowledge of God, so central in the tradition of Reformed theology. Such a premise, though, might lead one to think that belief in God mediated through other beliefs is an offense to God, perhaps inconsistent with God's omnipresence and immanence.

There is a sense, though, in which this metaphysical (theological?) basis for the inappropriateness of believing in God on the basis of reasons is closely tied to epistemological claims regarding the nature of a proper and well-formed noetic structure. In addition to allowing theistic belief a place in the foundations of a rational noetic structure, Plantinga sees Calvin making the further claim that "a Christian ought not believe in God on the basis of other propositions; a proper and well-formed noetic structure will in fact have belief in God among its foundations." Here the view that the Christian ought not believe in God on the basis of other rational beliefs stems from a particular view about where theistic belief should fit in a proper noetic structure, presumably where it should be according to a theistic design plan. It should be basic.

This point, repeated by Plantinga several times in "Reason and Belief in God," is stated most lucidly in the following:

Hence, we get something like:

Although Plantinga suggests this as in fact a position held by the likes of Calvin, Kuyper, Bavinck, and Barth, Plantinga makes another statement, a little less extravagant, when he talks about an (instrumental) use of natural theology as "a means of moving toward what Calvin sees as the best way to believe in God: as basic."¹⁴ Similarly, in an earlier article, Plantinga says that according to the Reformed tradition "the most appropriate way to believe in God is not to believe on the basis of evidence or argument from other propositions, but to take this belief--that there is such a person as God--as basic." Notice that in these two references, the proper basicality of belief in God is the "best" and "most appropriate" way to believe in God. This is different from saying that taking belief in God as basic is the correct and proper way to believe in God, suggesting that to do otherwise is inappropriate. Similarly, perhaps taking theistic belief as basic is more consistent with the types of metaphysical and theological points addressed above. But what would be the epistemic significance of such a noetic structure? Neither Calvin, Kuyper, nor Bavinck specify what relevant epistemic categories are involved in the distinction between basic and nonbasic belief in God. Plantinga, though, clearly wants to hold that basic theistic belief is (at least in general) epistemically superior to inferential theistic belief, and the key to this position lies in the idea that the immediate source(s) of theistic belief (generally) confer a greater degree of warrant than when theistic belief is based on reasons. This in turn may be traced to a belief that humans have been designed to form theistic belief in a basic way.

Plantinga, then, wants to say something like:

I want to distinguish, then, between two versions of RE. I will call Soft RE the position stated in [R2] and Hard RE the conjunction of [R2] and [R3] or [R4]. Is either Soft or Hard RE compatible with evidentialism? One way to show that one of these is compatible with evidentialism would be to find a theory that is compatible with one of the two and which also entails something in the neighbourhood of an evidentialist requirement. To this I now turn.

According to William Alston a belief that p is justified if and only if it is based upon an (externally) adequate (internally accessible) ground, where (a) the ground consists of experiences or other beliefs (or knowledge) of the subject, and (b) "adequacy" is understood in terms of truth-conducivity--given the ground, the belief that p is very likely to be true.¹⁶ Moreover, the structure of justified belief is foundationalist in character:

Accordingly, there is a distinction between immediately (or directly) justified beliefs (where what does the justifying does not include other beliefs) and mediately (or indirectly) justified beliefs (where what does the justifying includes other beliefs). Although the latter may depend for their justification on beliefs which are themselves mediately justified (and so on with the grounds of these beliefs), any given mediately justified belief will depend ultimately on a set of beliefs which are immediately justified. And the line of descent for any mediately justified belief will not be a single line of descent, but will generally depend on several beliefs, each of which in turn is based upon several beliefs, until the foundations are ultimately reached.

Hence,

One of the main reasons why philosophers have adopted some form of foundationalism is that it appears impossible to be mediately justified in a belief that p without the kind of structure articulated by the conjunction of [F1] and [F2]. More specifically, it seems that if S is to be mediately justified in any belief that p, the over all structure of justified belief must be such that S is immediately justified in some belief b as the terminus of any branch of justified beliefs which has issued from an original belief that p which is mediately justified. There are, the foundationalist argues, only four possibilities for any given branch of justified beliefs (issuing from a mediately justified belief): (a) it terminates in an immediately justified belief, (b) it terminates in an unjustified belief, (c) the belief that p occurs again at some point on the branch past the origin (and the branch forms a loop), or (d) the branch continues infinitely and has no terminus. The foundationalist argues that only if a branch assumes the form given in (a) will a necessary condition for being mediately justified in some belief that p be satisfied. The alternatives generate such difficulties as the infinite regress and circularity.¹⁸

Foundationalism, of course, has been charged with ending up in arbitrary dogmatism since it rests the structure of justified belief upon immediately justified foundations, and it is inferred from this that since such beliefs are immediately justified there can be no reasons in support of the foundations.

As Alston puts it in "Two Types of Foundationalism":

Alston has devoted several papers to articulating and defending a version of modest foundationalism which, among other things, answers this objection to foundationalism.²⁰ Alston's version of foundationalism distinguishes between various epistemic levels built up by the introduction and iteration of pistic and epistemic operators, thereby creating multiple levels of belief, justification, and knowledge.

Given any belief that p, we may distinguish between the belief that p and various higher-level doxastic correlates of p, the candidates for which would include S's belief that: (a) p is a rational belief, (b) p is justified (immediately or mediately), (c) p was formed in a reliable manner, or (d) p is based upon adequate grounds. Call such higher-level correlates: beliefs that P*. Among other things, this allows us to distinguish between a belief that p and a higher-level belief about the epistemic status of p. Let us call higher-level "epistemic" correlates: beliefs that P_j*, where P_j* = [p is justified]. By adding the subscripts "i" and "m" in parenthesis we can specify the mode of justification. Hence, the belief that P_j(i)* will be the belief that [p is immediately justified], and the belief that P_j(m)* will be the belief that [p is mediately justified].

One of the important consequences of this for foundationalism is that it makes possible (at least in principle) the assessment any belief, even an immediately justified belief, in terms of reasons, for even where a belief that p is immediately justified, it is possible (in principle) to find reasons for the higher-level belief that P_j(i)*. So, for example, where S's belief that [It is raining outside] is immediately justified, we may say that there exists a correlative higher-level proposition [<It is raining outside> is immediately justified]. Alston argues that, even if a putative belief that p is immediately justified, this does not preclude seeking reasons for p's correlative higher- level proposition [p is immediately justified]. Since every nonepistemic belief that p may be thought of as having a correlative epistemic belief that P_j*, even if one is restricted to immediate justification on the lower-level for p, one may adduce reasons at the higher-level for the belief that P_j(i)*. So, though S may be immediately justified in believing that [it is raining outside], S may have reasons for regarding the belief that [it is raining outside] as immediately justified. In other terms, S may have a mediately justified belief that [S's belief that <it is raining outside> is immediately justified]. Hence, though the belief that p will be immediately justified, the belief that P_j(i)* will be mediately justified.

Recalling Alston's concept of justification (a belief that p is justified if and only if it is based on an adequate ground), what the justification of any belief depends upon is the existence of a valid epistemic principle that applies to the belief. S need not (justifiably) believe or know anything about the justificatory conditions that obtain (though the ground--not its efficacy--will be cognitively accessible). To be justified in some belief that p requires (a) the existence of a valid epistemic principle and (b) p's satisfying the conditions the principle lays down. But it is precisely the subject's coming to have (justified) beliefs the content of which is given in (a) and (b) that will come to bear on the higher level. For (a) and (b) are the types of reasons we will have for regarding some belief that p as justified. Call these meta-reasons, as they are reasons for regarding some belief as justified. So whereas an immediately justified belief that p requires the existence of a valid epistemic principle that relates to the belief in question, a mediately justified belief that P_j(i)* will be based upon the (justified) beliefs that the epistemic principle for p is valid and that p may be subsumed under the principle.

Hence, we may add the following principles as constituents of Alston's foundationalism:

The principle of epistemic levels:

For every putative immediately justified belief that p, there is a higher-level evidential option:

But the position envisaged here is stronger than this. According to Alston, the justification of any higher-level belief would actually require mediate justification. This gives rise to what I will call the higher-level evidentialist requirement. Any belief that [p is justified] is, if justified, mediately justified. This is not to say that one can't be immediately justified in the lower-level belief that p without being mediately justified in a higher-level belief (a clear case of level confusion), it is only to say that if one is justified in any higher-level belief that [p is justified] then that belief must be based upon adequate reasons. Alston's position here rests on certain arguments against the plausibility of immediately justified epistemic beliefs. If we examine two main models of immediate justification, that of direct experiential awareness and self- evidence, neither seems plausible to uphold the immediate justification of attributions of justification to one's beliefs. We do not seem to ever be presented with epistemic facts or statuses, nor is it ever self-evident that we are justified in some belief that p, even if we often take it without question that we are so justified. To put it another way, a belief that p is justified in virtue of their being a valid epistemic principle which lays down the conditions of justification for the belief in question, called it b. A belief about the justification of b is an epistemic evaluation of b and involves the application of certain standards. So, although one may be immediately justified in some belief that p, if one is to be justified in the higher-level belief that [p is immediately justified], then one must justifiably believe that the principle relating to p's justification is valid and that p fulfils the conditions of p's epistemic principle.²¹ What arises at the higher- level is the validity of certain epistemic principles and matters of fact as to whether the original belief can be validly subsumed under the appropriate principle of justification. It seems dubious that we are ever just presented with such matters, much less that they are self-evident.²² Hence, the justification of higher-level beliefs will be mediated by other beliefs of the epistemic subject, from which follows an evidentialist requirement for the justification of all beliefs about the epistemic status of beliefs that p.²³

In "Two Types of Foundationalism," Alston summarizes the notion of epistemic level distinctions in the following terms:

Let us, therefore, lay down the principle of the higher-level evidentialist requirement:

What I want to now suggest is that the conjunction of [F1]-[F5] is compatible with both Soft and Hard RE, and that conjunct [F5] entails an epistemically adequate form of evidentialism.

The claims of RE are most often made within a foundationalist framework of justified belief, though a foundationalist framework of a non-classical kind which allows a broader range of basic beliefs (thereby allowing theistic belief a place in the foundations). I take it, then, that [R2]-[R4] are compatible with the foundationalism articulated by [F1] and [F2]. Moreover, I will take it as without controversy that [R2]- [R4] are compatible with the commitment to the existence of correlated higher-level beliefs established in [F3]. Hence, we may distinguish between the belief that Pt and such higher-level correlates as S's belief that Pt is justified, is reliably formed, etc. But the distinction between epistemic levels entails that there are actually two distinct levels on which reasons may be introduced. [F4] laid out an evidentialist option to adduce reasons for the correlated higher-level epistemic belief that P_j*, whereas [F5] gave us a stronger claim to an evidentialist requirement imposed on such higher-level beliefs. There is a lower-level evidentialism which requires a mediate justification of nonepistemic beliefs, the theistic evidentialist correlate of which is:

and there is a higher-level evidentialism ([F5]) which involves the imposition of an evidentialist requirement on the correlative epistemic beliefs of nonepistemic beliefs, which for theistic belief entails:

For my purposes I will focus on [F5] rather than [F4], as I am more interested in the prospects for compatibility between RE and an evidentialist requirement than the weaker (though not necessarily less interesting claim) that RE is compatible with offering reasons for higher-level epistemic beliefs.

As a first move I want to argue that we should reinterpret theistic evidentialism and the claims of RE in the light of Alston's multi-level foundationalism and its commitment to the higher-level evidentialist requirement as formulated in [F5]. As I have argued, [E] is incompatible with Reformed theology since it rules out immediately justified theistic beliefs. However, now that we have introduced higher-level beliefs, we can modify our previous conclusion. All that was established earlier was that, since--for the Reformed theologian--theistic belief is immediately justified for some people, basing such beliefs upon adequate evidence cannot be a necessary condition for its justification. Consequently, the evidentialist requirement was rejected with respect to lower-level beliefs. Recognizing this, there is a prima facie plausibility in holding that higher-level beliefs about the epistemic status of putative theistic beliefs always require adequate reasons for their justification. What we have found in the Reformed tradition is an emphasis upon immediately justified beliefs at the lower-level (e.g., God is speaking to me, forgiving me, is present). From this follows only the rejection of lower-level evidentialism. This leaves open the possibility of advocating evidentialism on the higher level with respect to all epistemic beliefs. So, since it is possible for the belief that Pt to be immediately justified even if the belief that Pt_j(i)* can only be mediately justified, [R1] and [R2] are compatible with [E*]. The crucial question, though, is whether either [R3] or [R4]. And this question would seem to rest on whether it is possible that the belief that Pt_j* be nonbasic in some noetic structure N and the belief that Pt be basic in N.

Consider a sensory perceptual case where John believes that he sees a tree and this belief is based upon some experience e, such as being appeared to treely. John's belief that he sees a tree is justified if and only if it is based upon an adequate ground. Let us assume that e is an adequate ground here. John then takes a class on epistemology taught by Alvin Plantinga and he reads William Alston's Epistemic Justification and unable to withstand Alston's powerful argumentation, he straightaway becomes a modest foundationalist of the Alstonian variety. Late in the term, while sitting at the table in his kitchen attempting to prepare the final term paper for Professor Plantinga's course, John looks out the window and, upon being appeared to treely, the belief [I see a tree] is formed. John recalls some of what he has learned during the term. He knows he has good reasons to believe that [I see a tree] is immediately justified for him at time t. He (justifiably) believes that the experiential ground of [I see a tree] is adequate and that it is rational to suppose that the sensory perceptual doxastic practice is a reliable mode of belief formation. On the basis of these (and like) consideration John believes that [<I see a tree> is immediately justified for me at time t]. This higher-level belief is nonbasic, but, even while meditating on this higher-level question, his eyes are fixed fast on the Old Oak tree just outside, and he believes, on the basis of being appeared to treely [I see a tree] and this belief is immediately justified for him.

What is essential to grasp in this example is the distinction between the supports- relation and the basis-relation in a noetic structure. A belief b will be basic if and only if b is not believed on the evidential basis of some other belief(s) c, and a belief b will be nonbasic just if b is based upon some other belief(s) c. However, a belief b may be basic in a noetic structure even if some other belief(s) c in that structure provides evidence for b. This would mean that S has propositional evidence for b, but S does not believe the relevant proposition on the evidential basis of this propositional evidence. Having some ground G₁ for a belief b is compatible with b's being based on some other ground G₂, and even with not believing b at all. The distinction between a belief's being based upon an adequate ground and a person's merely having an adequate ground for her belief is an important one. Having adequate grounds for believing that p (and hence having a justification) is independent of believing that p. I may have a set of adequate reasons for believing that I will not pass my Latin mid- term exam (e.g., I haven't studied all week, my last two exams were D's, and I was absent the day the instructor went over the material to be covered on the exam), yet I remain unmoved in my belief that I will pass it (perhaps as a case--all too common-- of wishful thinking). Here I have adequate reasons to believe not-p (I will not pass the exam), but in fact I believe p. But this independency suggests that having adequate reasons for the belief that Pt is actually compatible with believing Pt in a basic way. In other terms, the supports-relation in a noetic structure is not the same as the basis- relation. The belief that Pt can be basic in some noetic structure N, but nonetheless be supported by other propositional items found in N. S may believe that Pt in a basic way, but S may believe other things, such as "the universe exhibits design," which provide evidential support for Pt. If this is the case, though S doesn't believe that Pt on the basis of reasons, S may have reasons, which--if adequate--would entail that S has a mediate justification for believing Pt.

So even if the doxastic grounds upon which a higher-level belief B* is based provide reasons which support the lower-level correlate b (which surely seems to be the case), the existence of such evidence within S's noetic structure does not entail that b is nonbasic. It is both logically and psychologically possible (in fact quite common) for epistemic beliefs to be held on the basis of ratiocination though the lower-level correlate remains a basic belief for S. Thus there is no logical or psychological difficulty in maintaining the proper basicality of theistic belief and also holding that beliefs about the epistemic status of theistic belief are, if justified, mediately justified. It is possible for S to believe that Pt on the grounds of religious experience and yet for S, as a cognitively reflective person, to believe that Pt_j(i)* on the grounds of adequate reasons, perhaps stimulated by a reading of Plantinga, Wolterstorff, or Alston.²⁵ Consequently, [E*] is compatible with both [R3] and [R4], and therefore is compatible with Hard RE.

But how does the higher-level evidentialist requirement stand vis-a-vis evidentialism? Intuitions regarding what constitutes an adequate evidentialist requirement need some straightening out before considering how [E*] measures up to [E]. In the previous section I suggested that epistemic level distinctions had some important consequences for the evidentialist requirement for belief. With the background of level distinctions, the evidentialist requirement can no longer be looked at as merely ranging over some set of beliefs that p, where every p is of the 0^th level. We now have to contend with multiple levels 0^th, . . . ,N^th on which this requirement can (in principle) be imposed, and from what has been argued above the justification of every epistemic belief is in fact subject to such a requirement. More generally, for every belief that p, there is an evidentialist requirement for p's correlated higher-level proposition P_j*. Could it be, though, that such a requirement actually satisfies intuitions which have motivated traditional evidentialism without, at the same time, being saddled with the difficulties encountered in the traditional form? I think so.

I want to argue that the impulse leading to traditional evidentialism is the connection of epistemic justification (or at least what we call epistemic justification) with the notion of reflective rationality. Being justified is not merely a matter of the believing subject being in a certain good cognitive state vis-a-vis some proposition, but of the subject being aware that he actually is in that good cognitive state. And why is that important? If, as some writers have recently suggested, our idea of justification has developed within the social context of answering objections to our beliefs, the ability to ask and answer questions about the grounds of our beliefs and their adequacy will become quite important. And this easily leads to two possible consequences: (1) a maximal requirement that individuals themselves have good reasons for a belief, especially if that belief is the type of belief which is subject to doubt by large segments of society, and (2) a minimal requirement that there is evidence somewhere in the community for a belief. Another way of putting the matter is to say that traditional evidentialism is strongly motivated by the need to impose some kind of internalist constraint. Consider the case of merely reliable belief formation. Some person forms the belief that the Dallas Cowboys will win the 1994 Super Bowl because it was the result of some apparently irrational conviction by which the person is seized every year prior to the American Football season opener, but such convictions always turn out to be correct. If the mere reliability of the mode of belief formation were sufficient for justification, the belief in question would be justified. And is this not absurd? Intuitively there is a strong drive to add, "Well, if you can know that this mode of belief formation is reliable, or in some way become aware of this fact, then (and only then) the belief is justified." What S lacks in the original scenario is anything by way of grounds or reasons which would give S something to go on as a sign that the belief in question is true. So we are naturally led to hold that S's being justified in some belief that p consists not merely in the belief's being produced in a reliable fashion, but also that we have reason to believe that it was produced by a reliable process, or that the grounds are adequate, or something along these lines. In other terms, knowledge is important because having true beliefs is important, and justification is important because it places us in a good position with respect to possessing true beliefs. But this latter state of affairs is only as good as the ability of the individual to be "cognitively in tune" with the fact that he is in such a positive epistemic state.²⁶ And the most natural way of working this out is to impose an evidentialist requirement for belief, possibly exempting those cases of belief which are (allegedly) self-evident, incorrigible, or characterized by some kind of epistemic immunity.

Where traditional evidentialism goes wrong is in its failure to distinguish between reflective (second-order) and unreflective (first-order) rationality or justification. The goal of critical reflection on our beliefs is an important aim, but it should not be confused with the first-order goal of securing true beliefs, or at least of being in a strong position with respect to acquiring true beliefs. The Enlightenment view of man as typically critically reflective seems generally not to be the case for many (if not most) epistemic subjects. Pre-reflective children, idiots, and unreflective adults are not capable of carrying out a philosophical analysis of their beliefs, subjecting them to the types of questions the epistemologist will ask. And even where people are capable of doing so, they generally do not. Are we to deny to them the state of justified beliefs (or even of possessing knowledge) simply because they are not reflectively rational in a wide range of beliefs? What must be concluded is that at the lower-level people do acquire many rational beliefs without these beliefs being based on reasons, let alone on the basis of reasons regarding the adequacy of the grounds of the beliefs in question. Justified beliefs in the adequacy of the grounds of one's beliefs enter into the picture in a different way. They are required not for being justified in the belief that p, but for being justified in the belief that P_j*. They are not required for rational belief, but for reflectively rational belief. But the two cognitive desiderata must be distinguished.

We may say, then, that being justified in the higher-level belief that Pt_j* is necessary for reflective rationality in the belief that Pt. I say necessary, and not sufficient, because I am taking rationality to be truth-conducive. On this way of looking at things being justified in believing that [one is justified in believing that p] does not entail that one is justified in believing that p, for to be justified in believing that p is for one's belief that p to be based on adequate (i.e., truth-indicative) grounds, but S may justifiably believe that the grounds of the belief that p are adequate when in fact they are not. But if S's belief that p is reflectively (truth-conducively) rational, then it is (truth-conducively) rational. So being justified in the belief that Pt_j* is necessary and, together with unreflective rationality (at the lower-level), sufficient for reflectively rationality. If S is justified in believing that [S is justified in believing that p], but S is not actually justified in believing that p, we can say that S has an attenuated reflectively rational belief that p. Here S has adequate reasons for supposing that the grounds of the belief that Pt are adequate, though in fact they are not adequate.

In the light of [RR] we can also revive the epistemic significance of having reasons for belief, for if [RR] is true, then a person who is reflectively rational in believing that Pt_j* also has reasons for the lower-level belief that Pt--specifically S will have reasons for supposing that the grounds of the belief that Pt are adequate, reasons for supposing that the belief is truth-conducively justified. Now the two deficiencies of merely having reasons for a belief (as opposed to a belief's being based on reasons) are well-advertised in the literature: it is compatible with (i) not believing p and (ii) believing on the basis of an inadequate ground.²⁷ But if S's belief was in fact immediately justified (by being based upon an adequate ground), then if S also had reasons for Pt, then this would be epistemic icing on the cognitive cake. In such a case, S's having reasons for the belief that Pt is conjoined with S believing that Pt and the belief that Pt being based upon an adequate ground. Perhaps such a situation would be a (limiting) case of a belief's being partly immediately and partly mediately justified (where reasons a person has for a belief can add to its epistemization). And this is quite compatible with [R3] and [R4], for what they tell against is a noetic structure in which reasons constitute the sole source of justification for theistic belief. But there is always room for a noetic structure in which the belief that Pt is based on adequate (nonpropositional) grounds and yet S also has good reasons for the belief that Pt. Such a noetic structure could well be epistemically superior to an [R4] noetic structure.

Although traditional evidentialism has sought to satisfy the goal of reflective rationality by the general demand for adequate reasons for belief, I have argued that this demand is appropriately imposed for higher-level beliefs to the effect that one is justified or rational in some putative belief that p. In this section I have tried to show that the higher-level evidentialist requirement is an adequate form of evidentialism, and I believe I have shown that to follow from the satisfaction of intuitions of reflective rationality and internalism that have played a significant role in the development of both the concept of epistemic justification and the position of traditional evidentialism. So I conclude that [E*] entails an adequate form of evidentialism.

In this paper I have emphasized the relevance of the distinction between belief in God and belief in the rationality of belief in God for the prospects for the compatibility of RE and evidentialism. I have argued that Alstonian foundationalism, in which this distinction is highlighted, entails an evidentialist requirement for the latter, higher-level belief.²⁸ Moreover, since Soft and Hard RE are both compatible with Alstonian foundationalism, it follows that higher-level theistic evidentialism is a form of evidentialism that is compatible with Reformed epistemology.²⁹ That it is an epistemically adequate form of evidentialism I take to follow from its satisfying the cognitive goal of reflective rationality so much a part of the evidentialist tradition. I judge, then, that the prospects for the compatibility of RE and evidentialism are good indeed and worthy of further investigation.

Oriel College, Oxford

Many thanks to Robert Audi and Richard Swinburne for critical comment on earlier drafts of this paper. I am also considerably indebted to William Alston, Alvin Plantinga, and Nicholas Wolterstorff for their comments, criticisms, and suggestions regarding the analyses and arguments I have incorporated in the paper.