CHAPTER 1

Epistemological Prolegomenon

 

            Any account of religious epistemology (broadly speaking, an account of the justificatory or epistemic status of religious or theological propositions and/or beliefs) involves and to a large degree presupposes an array of issues in general epistemology. Since the central topic of this thesis will be the justificatory and epistemic significance of reasons or propositional evidence vis-à-vis theistic belief, this chapter will be devoted to (1) setting out, with some critical commentary, some of the basic concepts and distinctions in both substantive and meta-epistemology[1] which will be employed in the thesis and (2) articulating some of the general features of the epistemological framework to be developed in the chapters which follow with reference to theistic belief in particular.

 

I. Foundationalism

 

            One of the concerns of epistemology is to address the substantive question of the overall structure of justified belief (or knowledge).[2] We may think of our beliefs, together with their epistemic and psychological properties, as a complex network of doxastic relations. These doxastic relations together with the input of experience we can call a noetic structure. According to epistemic foundationalism, the doxastic and epistemic content of a person’s noetic structure divides into foundations and superstructure.

 

A. Justification and the Infinite Regress Problem

 

            To understand foundationalism it is helpful to begin with the sort of problem which motivates something like the foundationalist theory of justified belief.  I refer to what has been called the infinite regress problem. Here it is helpful to begin by thinking of a person’s noetic structure as composed of various justificatory chains - sequences of beliefs with (at least) the first belief being a justified belief, and where each belief in the sequence is linked to the previous one by being based upon it. Suppose now that S is justified in believing that p on the basis of some other belief q. In what sorts of justificatory chains could this scenario take place? There are four possibilities: S’s belief that p is justified by virtue of being based on some other (justified) belief q, where q is:  (1) a member of a finite chain of justified beliefs which at some point terminates with an unjustified belief B_n [finite and unanchored chain], (2) a member of a chain of justified beliefs whose doxastic sequence continues infinitely and has no belief B_n as a terminus [infinite and unanchored chain], (3) a member of a finite chain of justified beliefs which at some point in its doxastic sequence has a belief B_n that is based on the (original) belief that p (thereby forming a looped chain) [finite and redundantly anchored], or (4) a member of a finite chain of justified beliefs which at some point terminates with a justified belief B_n that is justified (solely) by something other than its relation to some other justified belief(s) [finite and anchored].

            This argument differs from what some have called the dialectical infinite regress argument (Audi 1993, p. 119) that develops in response to the question, “What justifies you in believing that p?”  This question is posed in the context of a skeptical challenge and is roughly equivalent to “Show me that you are justified in believing that p.” The argument here replaces the “p is based on q” clause above with “q is cited in support of p.” Such citings will be circular, involve an infinite regress, or terminate in a belief for which no justification is given since it is either unjustified (and so cannot be shown to be justified) or wears its justification on its sleeve (and so does not need to be shown to be justified). The regress argument I am considering is a structural one, resulting from a consideration (in a non-skeptical context) of either the entire body or particular item of justified belief possessed by any person. Here the question “How are you justified?” is not asked with skeptical force (in the sense of “Show me that you are justified”), but rather with informational force (in the sense of “How is it that you are justified?”).[3]

There is a sense in which the structural regress problem is more to the point than its dialectical counterpart. Whether or not there is a response to the skeptical challenge there still remains the question of what the structure of our justified beliefs is, what is required for the whole structure of our beliefs if we are to be justified in any belief.  Moreover, epistemic justification is properly a matter of a state a person is in vis-à-vis some proposition, as opposed to some activity of showing that one is in some positive epistemic state. The former seems to be the central concern of epistemology. Showing that a belief has some favourable epistemic status presupposes the condition of a person’s being in that favourable epistemic state, for the activity of showing aims at establishing that some positive epistemic status is actual for a person. In similar fashion, I may be in a having-received-a-letter-from-Bill-Alston state without showing that I have received such a letter. My being in that condition is presupposed in the activity of showing that I have received such a letter. For this reason I speak of the state as more fundamental than the activity.

            Two of the options surveyed above may be quickly ruled out. Since it is not possible for an unjustified belief to generate (or sustain) justification for other beliefs, we may eliminate (1). There simply cannot be any epistemic iustificatio ex niliho.  Nearly as dubious is the idea of any epistemic iustificatio ad infinitum (as (2) asserts). One reason is that such an infinite justificatory chain would appear to entail an infinite belief chain, and it seems psychologically (and perhaps logically) impossible for a human subject to have an infinite set of beliefs. Although what “having a belief” amounts to is crucially important to this, Audi (1993, p. 127) has argued that an infinite set of beliefs (say, arithmetical beliefs - 2 is twice 1, 4 is twice 2....) would ultimately involve a formulation too lengthy for a finite mind to understand. Even if we understood that the formulation expressed a truth, we would not understand the formulation as a whole. But if we cannot understand the formulation as a whole, we cannot grasp the truth it expresses as a whole, and if we cannot do that, we cannot believe the truth it expresses.[4]

            A bit more tricky is (3) - the circular justificatory chain. There is good reason to think that the based-upon relation between beliefs in a justificatory chain is at least in part causal. So if at time t₁ S’s belief that p is based on S’s belief that q, then S’s belief that q is in part causally responsible for S’s belief that p at time t₁. Hence, in a circular doxastic chain, at any given time t_n S’s belief that p would be caused (in part) by S’s belief that q, S’s belief that q would be caused (in part) by S’s belief that r, and S’s belief that r would be caused (in part) by S’s belief that p. If transitivity holds between causal relations (as most philosophers accept) and a belief’s being based on another belief is in part a matter of a belief being causally responsible for another belief, then the preceding entails that S’s belief that p is in part causally responsible for itself!  Notice that the kind of circularity involved here is synchronic doxastic circularity - circular relations between beliefs at a specific time and so causal sustenance at a specific time. This should be distinguished from diachronic doxastic circularity which involves something like [belief b₂ causally sustains b₁ and b₃ causally sustains b₂] at time t₁ and [b₂ causally sustains b₃ and b₁ causally sustains b₂] at time t_n+1. Of course, if the problem is generated by the causal entailments of the is-based-upon relation, one might modify the structure of the justificatory chain so that justification is a matter only of the evidential support relations between beliefs, not their causal grounding. The infinite regress argument has proceeded on the assumption that justificatory chains entail identical psychological chains. But perhaps at some point there is a divergence between what a belief is based upon and what evidentially supports the belief, where the latter is what sufficiently justifies the belief. The idea here is roughly this: the causal source of a belief may be nondoxastic, even if what justifies the belief is doxastic. The resultant justificatory circle is thus not a causal circle but an evidential circle. Such a circle does not entail beliefs that causally sustain themselves.

            It would be important though to spell out the notion of evidential support. Minimally this would involve something like a belief B₁ being evidentially dependent on B_n. On the assumption of the transitivity of evidential dependence, a circular evidential chain would involve (a) B₁ evidentially depending on some belief B_n for its justificatory status and at the same time (b) the belief B_n depending on B₁ for its justificatory status. Can this make sense? Suppose that this relation of evidential dependence is something like B₁'s being made probable by B_n.  (Specifically, I have in mind an evidential probability, the probability some proposition h has on some evidence e.[5]  More precisely, the kind of evidential probability here is what is commonly called epistemic probability, where the probability of h on e is relative to a human cognizer’s knowledge - of both logical and contingent evidence - and so can be increased or decreased by the addition of new evidence in the form of logical and contingent truths. I shall have more to say about evidential probability in II.D). The evidential circle would then amount to: B₁ is made evidentially probable by B_n and B_n is made evidentially probable by B₁. But if B₁ is not evidentially probable apart from B_n, how can B_n be made probable by B₁?[6]  However many links fall in between B₁ and B_n will not make any difference to the scheme. It simply is not clear how a belief is going to derive justification from a chain of justification that turns back on itself. Self-probabilification seems most dubious.  Moreover, if the making-probable relation is what confers justification, it looks like a circular justificatory chain will involve something very close to (at least partial) self-justification. But this would be one way to conceive of (4) and so would support some kind of foundationalist answer to the infinite regress problem.

            These circular justificatory chains amount to forms of linear coherentism. One way to escape the adverse consequences of these forms of linear coherentism is to combine the linear transference of justification in a doxastic sequence with a form of holistic justification. On this model the evidential support does not amount to mere linear relations between individual beliefs (though it may certainly include that). Rather evidence is ultimately a network of relations, and justification supervenes on the total set of beliefs the objects of which form a more or less comprehensive, coherent system of interconnected propositions. Coherence relations would typically involve the kind of large scale “fitting together” achieved by various entailment and explanatory relations between beliefs. So S’s belief that p is justified by virtue of being based on some (justified) belief that q, where q is (3*) a member of a finite chain of justified beliefs that at some point in its doxastic sequence has a belief B_n which is justified by virtue of being a member of the entire sequence that forms a finite set of coherent propositions. Or, alternatively, we can say that (3**) given any belief that p, S’s belief that p is justified by virtue of being a member of a finite set of coherent propositions.[7]

            Like linear coherentism, this holistic coherentism is what Pollock (1986) calls a “doxastic theory.” Justification is determined solely by a belief's relation to other beliefs, not the relation of beliefs to something else outside the mental life of the epistemic subject. S’s belief that p will be justified just if S’s belief that p exhibits the type of coherence just mentioned with the set A of S’s beliefs (q, r, s, and t). But sets of obviously false but internally coherent beliefs turn out to be justified on this account. Moreover, suppose set A is coherent, so will be set B which consists of the negation of the beliefs in set A. S₁ believes (p, q, r, s, and t); S₂ believes (not-p, not-q, not-r, not-s, and not-t).  Both sets are (or at least may be) equally coherent and so their doxastic contents justified. This may not strike one as problematic in itself, but consider: the doxastic components of A and B will be equally justified given the same set of nonpropositional empirical evidence which make up at least part of the experiences of S₁ and S₂. But something seems terribly wrong with this.  The intuition is that justification is not simply a matter of how beliefs cohere with each other, but of how well they fit with something outside the believer.  The problem raised here is that of epistemic isolationism or detachment from reality.[8]  A person’s noetic structure may fulfill the conditions of internal coherence but cohere very little with his experience at some time t_n (or indeed at all times). The experience of being appeared to treely at time t may not cohere with one’s system of beliefs or some subsystem within the system (perhaps one’s system or subsystem entails that no one is ever appeared to treely). From which it would follow that one is not justified in the belief that one sees a tree.  Suppose justification means or at least entails a making probable relation (as explained earlier). Is my belief that I see a tree or that I have a pain in my leg improbable because it does not cohere with my system of beliefs? There is the sense that many instances of sensory perceptual and introspective beliefs would be justified (probable) and the beliefs that entailed their negations unjustified (improbable) even if they do not cohere with the rest of one’s beliefs. As long as we grant that there is no necessary connection between doxastic coherence relations and conformity to a person’s subjective nonpropositional sensory and perceptual states, we are open to there being a significant discontinuity between one’s beliefs and the totality of one’s empirical evidence. From which it follows that if holistic coherentism is true, a person can be epistemically justified in believing a contingent empirical proposition even if that proposition is either incompatible with or rendered improbable by the totality of one’s empirical evidence. Conversely, it will be possible for one to be unjustified in believing many contingent empirical propositions should they fail to cohere with one’s doxastic system even if they should be rendered probable by one’s total empirical evidence. Surely this is mistaken.

 

B. The Thesis of Foundationalism

 

            The infinite regress problem draws attention to the epistemological difficulty generated by restricting the mode of justification for our beliefs to other beliefs (or knowledge), where these propositional items are either cited (as in the dialectical version) or simply function as the grounds of our belief (as in the structural version), or perhaps merely provide evidential support (to account for the holistic addendum). If we assume that S is justified in believing that p only if S’s belief that p is based on some other justified belief(s) of S, then we are going to be saddled with one of the three sorts of implausible  justificatory chains just considered. The foundationalist argues that only if the justificatory chain assumes the form given in (4) will a necessary condition for being mediately justified in some belief that p be satisfied. Justification requires a chain that is both finite and (noncircularly) anchored. Therefore, if S is justified in the belief that p, S must be justified in a finite set of beliefs which terminate (via the is-based-upon relation) in justified beliefs (or knowledge) which are justified (or known) without being based on other justified beliefs or knowledge.[9]  Foundationalism is committed to the fourth justificatory chain which is finite and anchored, but not subject to the justificatory liabilities of the other three we have looked at here. Central to the foundationalist case is the idea of immediately justified beliefs, beliefs whose justification does not require mediation through other beliefs. To be more precise, we should say that it is wholly immediate justification that is at stake: there being conditions sufficient for the justification of some belief(s) that does not include any other beliefs. Putative immediate justifiers would include:  (a) immediate experience of what the belief is about, (b) facts about the origin of the belief, and (c) in some cases, the mere fact that it is understood or believed (or simply, the truth of the belief). 

In contrast to this, mediate justifiers would include:  (a) having adequate propositional evidence for the belief in question, (b) arriving at the belief by valid deductive or good inductive inference, or  (c) basing one's belief on adequate propositional evidence, but where no explicit inference is involved.  In these cases of mediate justification, a belief is justified (solely) by virtue of its relation to other (justified) beliefs or knowledge.[10]  The foundationalist further claims that a noetic structure will have both mediately and immediately justified beliefs, and that the former will terminate (via the is-based-on relation) proximately or remotely in at least some beliefs that are immediately justified.[11]   More precisely, although some mediately justified beliefs may depend for their justification on beliefs which are themselves mediately justified (and so on with the grounds of these beliefs), any mediately justified belief will depend ultimately on a set of beliefs which are immediately justified. To which we might add, the line of descent for any mediately justified belief will not ordinarily be a single line of descent, but will typically depend on several beliefs, each of which in turn is based upon several beliefs, until the foundations are ultimately reached.

            But we need some clarifications and refinements.

            Although I have assumed that the is-based-upon relation involves some relation of causal dependence, a further spelling out of that relation is problematic (Pappas 1981; Pollock 1986, p. 37; Alston 1989c, pp. 227-229; Audi 1993, pp. 53-55). If John’s belief that it rained last night is based on his belief that the street is wet, then John holds the former because he holds the latter, and John’s holding the latter belief explains his holding the former belief. A reason q for which John believes p will be in part a reason why he believes p. So q will play some role in causally producing or sustaining p.  But because John’s holding this belief will be causally dependent on a number of things unlike the basis of his belief (e.g., the physiological state of his brain), it will be necessary to distinguish between the basis relation and other sorts of causal dependence. This has led some epistemologists to add something like a connecting belief requirement, according to which S believes or takes it that q supports p (Audi 1993, p. 21). This would surely seem to be supported by cases of maximally explicit inference, such as where John comes to believe that <it rained last night> by inferring this from a body of (propositional) evidence which he takes to provide adequate support for the belief in question. The belief-forming process seems to be guided by beliefs about an adequate support relation. Problems develop with cases of less than explicit inference, leading some to emphasize the nonoccurrent nature of the connecting belief (Audi 1993, p. 21)[12] and others to speak about a certain unconscious “taking account” of the relevant features of the experience or belief and the belief being formed “in the light of them” (Alston 1989c, p. 229). Slightly more controversial is Audi’s further suggestion that because of the connection with justification and knowledge, the basis relation must be construed as (partly) epistemic. If S₁ and S₂ both have good evidence, q, for the belief that p, but S₁ believes that p on the basis of q and S₂ believes on some other basis r (which is not good evidence for p), then - despite the fact that S₁ and S₂ both have the same evidence - what may render S₁ justified and S₂ unjustified in the belief that p is to be found in the way of the basis relation. And it might also be thought that if S’s belief that p is wholly based on his belief that q, then S is disposed to adduce q in trying to justify his belief that p (what Audi calls the subjective justification requirement).

            A distinction lurking in the shadows here is that between matters of evidential support and matters of the causal origin of a belief.  Typically, the foundationalist takes the immediacy and mediacy of beliefs to be both psychological and epistemic. There is first the claim that some beliefs will be psychologically immediate (or direct) and hence have no psychological intermediary. They are not held or believed on the basis of some other belief(s). Then there is the claim that some will be epistemically direct and hence have no epistemic intermediaries. They do not depend on some other (justified) belief(s) for their justificatory status. I noted earlier that one could admit the psychological basicality of some beliefs but deny their epistemic basicality, so holistic epistemic coherentism is compatible with some form of psychological foundationalism.[13]  Alternatively, one might endorse epistemic foundationalism but deny psychological foundationalism. This will be decided in part by whether one holds to a source-relevant concept of epistemic justification according to which the psychological source (or causal origin) of a belief is involved in a belief’s epistemic status. If so, there will be a convergence of psychological and epistemic foundationalism. What this suggests is that foundationalism is a flexible account of the structure of justified belief (and knowledge), one that may accommodate some coherentist conditions. A foundationalist noetic structure, then, will be an assemblage of experiences and beliefs (with varying degrees of firmness) together with their psychological and epistemic properties and corresponding modes of immediacy and mediacy.

 

C. Strong and Modest Foundationalism

 

            There are two general forms of foundationalism that should be distinguished. There is first what has been labeled Modern Classical Foundationalism - the foundationalist structure of justified belief (or knowledge) which has its origins in Descartes and the Enlightenment philosophers (Locke and Hume principally) and which has, in various forms, dominated the Western intellectual tradition to the 20th century.[14]  Alvin Plantinga (1983a, pp. 58-59) has distinguished between ancient/medieval and modern classical foundationalism. According to the former, foundational or properly basic beliefs are either self-evident (e.g., 2 + 2 = 4) or evident to the senses (e.g., it is raining outside). The latter retains self-evident propositions as foundational but usually substitutes incorrigible propositions for those which are evident to the senses, where incorrigible propositions are - roughly speaking - those whose content is about one’s immediate mental experience (e.g., being appeared to rainly).  Although some writers use “classical foundationalism” to refer to what Plantinga calls “modern classical foundationalism,” I will speak specifically of modern classical foundationalism (hereafter modern CF).  The main differentiating characteristic of modern CF is the sorts of beliefs it takes as proper for the foundations. More precisely, it employs criteria (or a criterion) which restrict the class of properly basic beliefs to those that possess any one of a number of epistemic immunities: (a) immunity to refutation (incorrigibility), (b) immunity to error (infallibility), or (c) immunity to the possibility of doubt (indubitability).[15]  As for the mode of support between foundational and superstructure beliefs, modern classical foundationalists have differed. Descartes, for instance, recognized only intuition and deduction as the appropriate paths to knowledge. Others have not been so stringent. Although Locke at points emphasizes deductive inference, he also seems to have countenanced (in Book IV of the Essay Concerning Human Understanding) what we might call broadly inductive, or probabilistic, inferences as being an adequate support relation between the foundations and superstructure.

            In contrast to modern CF, there is another form of foundationalism that derives from the Scottish common-sense philosopher Thomas Reid and has experienced a revival in the latter half of the 20th century under such philosophers as D.M. Armstrong, Anthony Quinton, Roderick Chisholm, William Alston, Alvin Plantinga, and Robert Audi. Because of its connection with Reid it is sometimes called Reidian foundationalism. Because of its contrast with modern CF, it is frequently referred to as “moderate,” “modest,” “minimal,” or “fallibilistic” foundationalism.  As these latter designations imply, Reidian foundationalism stands in contrast to “strong,” “radical,” or “extreme” elements in modern CF.  First and most importantly, Reidian foundationalism does not restrict the class of immediately justified beliefs to those that are infallible, indubitable, incorrigible, or in some way characterized by epistemic immunities. Candidates for immediately justified beliefs would include (a) cases in which S has an immediate experience of what the belief is about, (b) the truth of the belief, or (c) facts about the origin of the belief, and where none of these candidates for immediate justification need possess epistemic immunities. This opens up a whole range of foundational beliefs: memory beliefs, sensory perceptual beliefs, belief in other minds, various a priori beliefs, and possibly even moral and religious beliefs. Foundationalist theories of this modest variety can and in fact do differ as to what sorts of specific beliefs they allow within the foundations. Among other things, moral and religious beliefs are not universally taken as properly basic, even where epistemic immunities are not required. Secondly, although modern classical foundationalists differed on what sort of support-relations they thought appropriate, modest foundationalists have extended this as well. Candidates for mediately justified beliefs would include cases in which (a) S has adequate evidence for her belief, (b) S's belief is based upon some other beliefs, and (c) S arrives at a belief by means of inference (of some sort).

            To say that classical foundationalism has fallen on evil days would be a radical understatement.  The second half of the 20th century has witnessed some rather devastating criticisms of various aspects of modern CF.  Among such criticisms (to be explicated in detail in chapter 2) are self-referential incoherence and inconsistency with most paradigmatic properly basic beliefs.  But the failure of those foundationalist epistemologies that have taken Cartesian certainty as the criterion for foundational beliefs does not constitute a good argument against foundationalism as such. Of course modest foundationalism faces criticisms of its own, the primary ones being the related charges of arbitrary dogmatism and the epistemic independence of basic beliefs from the rest of a person’s beliefs. Foundationalism is often 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.[16]   Alternatively, there appears something disquieting about a person having justified beliefs without being able to show that they are justified. One may, it seems, claim for any belief one has that it is justified, noting that one may be justified without showing that one is justified. And if foundational beliefs are not certain, the foundations seem problematic. A good portion of the present thesis will be devoted to answering these kinds of objections with reference to one of the more controversial candidates for immediate or foundational justification, belief in God.  And what is true in the case of theistic belief is a consequence of more general difficulties with foundationalism as such.

            The foundationalist naturally holds that the foundations are justified independent of other beliefs or knowledge. But the notion of epistemic independence is susceptible to two different interpretations: (1) S is immediately justified in the belief that p = df. at time t₁ S has at least one justified corrigible belief that p whose justification does not at t₁ depend on (i.e., derive from) any other justified belief(s) of S and (2) S is immediately justified in the belief that p = df. S has a justified corrigible belief that p which would remain justified (ceteris paribus) regardless of any other beliefs S forms. Audi has distinguished these two in terms of momentary and omnitemporal justification (1993, pp. 106-107; cf. 159). So the notion of foundational beliefs does not imply that the same beliefs are always foundational, nor does their epistemic independence from other beliefs entail that they indefeasible. Omnitemporal justification may in fact be the result of thinking of justification as a relation between (timeless) propositions, whereas if it is a relation between a person’s beliefs then, given that a person’s beliefs change over time, the foundations and superstructure will not be invariant. Once the criterion of Cartesian certainty is removed from the picture, a modest foundationalist need not be committed to the indefeasible justification of foundational beliefs. He may admit that the justification for basic beliefs is susceptible to being overridden by sufficient reasons to the contrary, where these reasons are either drawn from the foundations or from the superstructure. So a foundationalist structure of justified belief may be modified from either the bottom up or top down.

            The charge of dogmatism may be answered in another fashion. The foundationalist need not deny that his position on being immediately justified in some belief that p rules out there being reasons for p.  First, a belief's being immediately justified for some person S₁ at time t₁ does not entail that it cannot also be mediately justified for some other person S₂ at time t₁, or for S₁ at some other time t₂. A belief’s being directly justified is both temporally and person relative. Secondly, if some particular belief that p is immediately justified, this does not preclude seeking reasons for regarding the belief as immediately justified. In short, there is no need to be committed to the thesis of iterative foundationalism: if S is immediately justified in believing that p, then S is immediately justified in believing that <S is immediately justified in believing that p>.[17]  Nothing in foundationalism as such need preclude the assessment of any belief, even a solely immediately justified belief, in terms of reasons. For even where a belief that p is immediately justified, it is possible (at least in principle) to find reasons for the higher-level belief that <S’s belief that p is justified>.[18]  Moreover, as I shall develop in chapters 6 and 7, it may be that one is justified in such second-order beliefs only if one has reasons for them. So that the apparent dogmatism that looms for immediately justified beliefs at the lower level is sufficiently answered by virtue of the fact that propositions at the higher-level are justified only on the basis of reasons.

            Another objection to modest foundationalism develops from the notion that there are no immediately justified beliefs since all beliefs are theory-laden or presuppose other beliefs. So for instance, sensory perceptual beliefs presuppose beliefs about the physical environment, spatial location of the subject and object, etc. The sorts of beliefs the opponent has in mind here may be appropriately labeled background beliefs. If we suppose that background beliefs do factor into the basis of some (or all) of one’s putatively basic beliefs, this only implies that in such cases the psychological source of the belief is partly propositional. Depending on how thinly we cut our basis and supports relations, there will always be the distinction between what is causally responsible for our belief and what justifies the belief, makes it probable, or evidentially supports it. The point is that the foundationalist can, on worst case scenario, assert purely immediately justified beliefs whose causal source is at least partly propositional. There can be an epistemic foundationalism without a psychological foundationalism. But it is not clear that background beliefs always factor into the psychological basis of putative basic beliefs. True, such beliefs are preconditions for our having many beliefs, but a precondition for having a belief is not necessarily causally operative (in the sense required by the basis relation) in the formation or sustenance of belief, let alone responsible for a belief’s being justified.[19] It would appear that even if some beliefs are at least only partly basic, others are wholly psychologically basic. Perhaps some of these beliefs require the possession of other beliefs and others merely a certain level of conceptual development. This does not preclude in any obvious way such beliefs being wholly psychologically (and epistemically) basic.

            Modest foundationalism, then, is able to answer some of the leading objections against it. Faced with the alternatives of coherentism and skepticism, the appeal of modest foundationalism should be evident.

 

II. The Concept of Epistemic Justification

 

            Well and fine - a foundationalist structure of justified belief. But what is justification? Perhaps we should begin with knowledge. Epistemology is concerned with knowledge.  More precisely, it is concerned with propositional knowledge, knowledge that such and such is the case, as opposed to knowledge how (to do something). Epistemology is interested in the conditions that must obtain if a person S is to have knowledge that p. What are these conditions? Nearly noncontroversial is the truth condition: if S knows that p, then p is true.  Slightly less noncontroversial is the belief condition: if S knows that p, then S believes that p (perhaps to some minimal degree).[20] True belief is, therefore, taken as necessary for knowledge. But is it sufficient?

            Since Plato[21] philosophers have made a distinction between knowledge and true belief. A person who by a hunch, coincidence, or lucky guess comes up with a right answer is not in a knowledge state with respect to a particular propositional object. On the way to the football game Sam tells Jerry that the home team will win. In fact they do. At the end of the game Sam turns to Jerry and says, “See, I knew they would win.” Sam did indeed hold a belief that turned out to be true. But did he know it? Or was his true belief an epistemic accident?  Plato's account, as well as most contemporary discussions, suggests that there is a surplus value of knowledge over true belief. To clarify what this surplus value amounts to  it might be said that what is necessary for knowledge is not only “getting it right” so to speak, but being situated such that one is in cognitive possession of that which functions as evidence or an indication of the truth of the belief, and so “getting it right” for the right reasons. In other terms, it is important that our beliefs not only be true but be grounded or supported in the right manner. And why is this? We desire knowledge because we want true belief, but perhaps the best way to guarantee that we have knowledge is to have beliefs for which we have something in the way of truth-indicating grounds, adequate reasons, or evidence. So a third element in knowledge may be seen as directly contributing to an epistemically good end. This third element is called “justification” (sometimes “warrant”) and during the second half of this century the orthodox formulation of knowledge has been justified true belief. [22]

 

A.  Preliminaries

 

            There is an initial caution that should be exercized in the treatment of the concept of epistemic justification. The locution S’s belief that p is justified possesses an ambiguity which easily gives rise to a rather basic confusion regarding the concept of epistemic justification. The ambiguity falls on the “is justified” component of the locution. S’s belief that p is justified may refer to a process that someone (perhaps S) has gone through (or perhaps an up-shot thereof) in order to show that the belief in question has some favourable epistemic status. It may also be taken to refer to a belief's possessing the property of a favourable epistemic status. Since “justified” or “justification” is often used to refer to the favourable epistemic status itself and sometimes to the process (or up-shot thereof) of showing a belief to have this status confusion easily results.  Distinguishing between a belief’s being justified and showing (or exhibiting) that a belief has the property of being justified is by no means trivial, for one’s concept of epistemic justification is significantly affected by whether or not one makes the distinction. Many accounts of epistemic justification simply fail to distinguish between the state or condition a person is in vis-à-vis some belief that p from the activity of producing an argument or showing that they are in the state. These accounts are vitiated by a confusion that frequently reduces the plausibility of the account of justification presented (or the criticism of some account of justification). Among such errors would be the rejection of immediately justified beliefs (and hence foundationalism), or at least the rejection of wholly immediately justified beliefs. If being justified involves justifying one’s belief (to oneself or another) reasons of some sort will have to be adduced in the form of other beliefs or knowledge. But if every justified belief is justified by virtue of mediation through other (justified) beliefs or knowledge, there can be no immediately justified beliefs. Consequently, I shall be taking “is justified” to refer to the state of being justified, what certainly seems to be the more fundamental notion and the concept which is central to epistemic justification. I shall be regarding “is shown to be justified” to refer to what I take to be a different concept altogether.

            Justification applies primarily to a belief or a cognitive subject’s having a belief, where this may be expressed as “S's belief that p is justified” or “S is justified in believing that p,” equivalent expressions indicating (i) S believes that p (though not necessarily consciously or occurrently) and (ii) the belief that p is justified for S.  It is only in a secondary and derivative sense that we speak of propositions being justified. Also, justification is an evaluative concept, and in this sense is contrasted with the “factual.” As William Alston has stated: “To say that S is justified in believing that p is to imply that there is something all right, satisfactory, in accord with the ways things should be, about the fact that S believes that p” (1989c, p. 83). A justified belief, then, will have a positive epistemic status. Moreover, as intimated, justification is concerned with the epistemic dimension of evaluation, where the epistemic dimension has regard for the truth goal of our believings - the aim of believing what is true and not believing what is false (which for some translates into a broader goal of maximizing truth and minimizing falsity in the corpus of one’s beliefs). To say of a belief that it is epistemically justified is to say that it receives high marks relative to this epistemic point of view. Fourthly, justification comes in degrees and is both time and person relative. This follows in part from the first feature of justification. As far as degrees of justification is concerned, it should be clear that if a belief is justified by virtue of some evidence (whether propositional or nonpropositional), the degree to which the belief  is justified will be partly a function of the amount and/or strength of evidence.  Most concepts of justification will have considerable common ground on the preceding points. What generates the Janus-faced character of justification is the different ways philosophers think that conditions can be desirable or commendable from the epistemic point of view, the different ways we can think of justification as counting towards the truth of a proposition.

 

B. The Deontological Concept

 

            One such perspective on justification is generated by thinking of beliefs in a way analogous to actions, as subject to obligations and duties.[23] To take the relevant analogy from human behaviour, Billy was justified (morally, prudentially, or legally) in doing A just if in doing A Billy was not in violation of any relevant duties or obligations. This is not to say, of course, that Billy was obligated or required to do A, only that he was permitted to do A (i.e., that A’s negation was not obligatory), because his doing A did not involve him in the violation of any relevant rules or regulations. So if Billy is legally justified in purchasing alcohol at age 21 this means that Billy does not contravene any of the laws of the state in purchasing alcohol, not that there is a law which places an (legal) obligation on Billy to buy a bottle of Jack Daniels. More precisely, as Alston (1989c) has suggested: “We may think of requirement, prohibition, and permission as the basic deontological terms, with obligation and duty as species of requirement, and with responsibility, blameworthiness, reproach, praiseworthiness, merit, being in the clear, and the like as normative consequences of an agent’s situation with respect to what is required, prohibited, or permitted” (p. 115).

            The paradigm cases of justification in the area of human behaviour serve to illuminate the evaluation of beliefs in terms of duty, obligation, and the like. S is justified in believing that p just if there are no relevant rules or principles that would prohibit S's believing that p.  As in the case of moral or legal justification, S’s being justified in believing that p does not mean that S had an obligation to believe that p (though this might in fact be true), only that the relevant rules did not prohibit believing that p.  But what sorts of rules or regulations can govern beliefs?  Since we are talking about the “epistemic” justification of belief (as opposed to say, prudential justification of belief), the relevant concept will be that of not violating any epistemic, cognitive, or intellectual obligations. If we return to the notion of the epistemic point of view, we will discover that acquiring true beliefs and avoiding false beliefs serves as the guide for determining epistemic duties and obligations. Presumably we will have duties like refraining from believing in the absence of sufficient evidence (whatever evidence amounts to) or accepting those propositions which we see to follow from (or to be rendered probable by) some other proposition(s). As Alston explains: “On a deontological conception of justification, the principles will forbid beliefs formed in such a way as to be likely to be false, and either permit or require beliefs formed in such a way as to be likely to be true” (1989c, p. 117).  Clearly, the deontological conception is subject to several permutations, depending on how exactly we spell out the relevant set of epistemic duties.

            Of course if justification is construed deontologically, then it would seem that we must have direct voluntary control over our beliefs, to be able to take them up “at will.” Otherwise concepts such as obligation, duty, reproach, and blame would not be applicable to them. Such is the case at any rate if we assume the wildly endorsed, but not uncontroversial, “ought implies can” thesis. Since the direct voluntary control of our beliefs (so-called direct doxastic voluntarism) is generally regarded as implausible, some epistemologists (Alston 1989c) have abandoned justificatory deontologism. This would include beliefs that are obviously false (e.g., the moon is made of green cheese), obviously true (e.g., 2 + 4 = 6), or somewhere in between there two extremes (e.g., religious, philosophical, or higher-level scientific propositions).  One very important consideration here (introduced by Swinburne 1981, pp. 25-26) is a conceptual one. Our reason for trusting our beliefs is the conviction that they are formed in us by factors independent of our will. If a person were to control the formation of his belief directly, say that there is a tree in front of him, he would know that this belief originated from his will. But then he would know that his belief originated independent of whether what it reported was the case and so had no actual connection to whether there was a tree in front of him. It is hard to see how the person could actually hold the belief if he knew, in effect, that he had no reason to trust the deliverances of his senses. There are clearly cases of apparent choosing to believe propositions at will: the making of an assumption for practical or theoretical purposes and acting as if p is true without actually believing that p, seeking to bring oneself to believe that p, asserting that p, or aligning oneself with a group committed to the belief that p.  But none of these voluntary activities should be confused with the state of believing that p.[24]

            To rule out direct voluntary control, though, is not to rule out indirect voluntary influence (Alston 1989c, pp. 134-142). We may embark upon various sorts of research, investigation, or study which puts us into contact with evidence of “this” or “that” sort, knowing that this activity may have certain negative or positive doxastic influences. We voluntarily engage in practices which will influence the formation of new beliefs, withholding belief, or the rejection of old beliefs, but these doxastic states each constitute an involuntary mental response to the evidence before us at any given time. Accordingly, a deontologist might modify the nature of epistemic obligations to account for this. Instead of attaching obligations, permission, blame, etc. to beliefs, they would be applied to those practices (within our control) which influence factors that in turn influence belief. So, for example, the relevant obligation might be something like doing as much as can be reasonably expected of one to see to it that one believes a proposition only if one has adequate evidence for it. But to the extent that the criticisms above stand it does not seem that obligations can attach to individual beliefs.

 

C. NonDeontological Concepts: The Reliabilist Constraint

 

            But even with this modification, the deontological concept may still be thought to possess an important liability, a liability from the epistemic point of view. Suppose we think of justification as a matter of being in a good or strong position to acquire a true belief. This kind of “truth-conducive” believing naturally translates into something like “it being likely that the belief that p is true,” where likelihood of truth is some kind of objective probability (see below). In other words, suppose that justification requires reliabilism of some sort.  Suppose that there is some person S who is a member of a primitive community and has lived all his life in this community on a remote island. S accepts the traditions handed down from the elders. S has never encountered anyone, or even heard of anyone, who has questioned the traditions of his tribe. Some of the traditions of the tribe describe events which (a) took place two generations past and on another, local island and (b) for which S is not in a position to gather independent evidence. S forms an array of beliefs on the basis of the traditions. Could S be held blameworthy, or in dereliction of an epistemic duty, for this? He has done all that could be reasonably expected of him in the conduct of his epistemic life. His grounds for the beliefs he forms are, as far as he can see, the best he and others in his community have for their beliefs. But if we suppose further that the traditions have not be formed in a reliable fashion, so that they are not a reliable basis for beliefs about the respective events, then S may in fact be deontologically justified but not truth-conducively justified. Equally, there may be a truth-conducive justification without deontological justification. Suppose that S₁ forms the belief that p one the basis of some other person S₂'s testimony. If S₁ had checked into the credentials of S₂ (as S₁ should have done in the situation), S₁ would have found evidence which suggests that S₂’s word is not to be trusted. But this evidence was in fact misleading and S’s testimony is highly reliable. Here, if S₁ had done his epistemic duty he would not have trusted the ground of his belief that p and not formed the belief that p. He would have been deontologically but not truth-conducively justified.[25]

            Favouring some kind of reliability constraint, a number of epistemologists construe justification in nondeontological terms (Swain 1981 and Alston 1989c), or simply drop justification talk altogether in favour of reliability (Dretske 1981 and Nozick 1981). I will focus on reliability-justification theories. Rather than think of justification as involving notions such as obligation and permissibility, some philosophers prefer to think of it in evaluative terms like goodness, desirability, and preference, and then associate these with the strong epistemic position by adopting some kind of reliability theory of justification.[26]  As Frederick Schmitt explains: “The idea that justified belief is belief that contributes to the end of true belief is most straightforwardly developed by identifying it with reliable belief - belief of a sort that is generally true” (1992, p. 2). One way of developing this is a reliable process theory, according to which a belief’s being produced by a reliable belief-forming mechanism or process is necessary and sufficient for the belief’s being epistemically justified. For instance, Alvin Goldman writes: “The justificational status of a belief is a function of the reliability of the process or processes that cause it, where (as a first approximation) reliability consists in the tendency of a process to produce beliefs that are true rather than false” (1979, p. 10).  Ernest Sosa states the generic or paradigm reliabilism (developed by philosophers such as Goldman, Dretske,  Nozick, and Swain) as follows:  S's belief that p at t is justified iff it is the outcome of a process of belief acquisition or retention which is reliable, or leads to a sufficiently high preponderance of true beliefs over false beliefs (Sosa 1991, p. 131).  There are a number of ways of unpacking process-reliabilism. The reliability of a belief-forming process may be taken as a function either of the number of true beliefs or the ratio of true beliefs to false ones. There are also different versions of reliabilism depending on the sorts of situations in which reliability is justificationally operative. A belief-forming process may be justificationally operative if and only if the process is reliable in (a) the actual world (actualist reliabilism), (b) all worlds having the same causal laws as the actual world (causal reliabilism), (c) all possible worlds (unrestricted reliabilism), or (d) all worlds consistent with our general beliefs about (objects, events, and changes in) the actual world.[27]

            But problems arise. Given reliabilism, beliefs that just pop into a person’s head will be justified (if they are the output of a reliable process) even if the person knows nothing of the mechanism in question or anything in the way of grounds for the belief, or on some accounts even if the person has good reasons for thinking that the belief was formed in an unreliable fashion.[28] Alternatively, if a person’s belief is the output of an unreliable process (e.g., one’s sensory perceptual beliefs are generated by a computer to which one’s brain is connected while it floats in vat) it is unjustified.  This will be true even if the person has no reason to believe that his belief is being unreliably generated, indeed even if the person has considerable evidence that it has been reliably generated. There seems to be no way for the reliabilist to distinguish between the reasonable and unreasonable epistemic behaviour of epistemic subjects who are in a vat. There will be no difference between a vat-man who makes a judgement about his sensory environment under conditions that appear normal to him and a vat-man who makes the same judgement under conditions that he believes to be abnormal and are creating the illusion that things are such and such. Yet another objection, the so-called generality problem, arises in relation to some accounts of what the relevant process is which produces justification. Where the process is a process type, since any concrete belief-forming process will be a token of many different types, all with varying degrees of reliability, the selection of which is the relevant type will determine whether the process is reliable. If the selected type is too narrow (e.g., perceptual beliefs resulting from examining brown objects which are less than five feet away under bright lighting, etc.), it is all too easy - barring counterfactual constraints - for it to be reliable (for every true belief can be described as the output of a very specific process). If the type is too broadly described (e.g., vision, hearing,), it will produce reliable beliefs as well as unreliable beliefs. 

            These problems suggest that reliabilism is in conflict with a widely shared intuition that a subject’s perspective or viewpoint (broadly understood) is relevant to justification. This difficulty may be alleviated partly by replacing reliabilism with a reliabilist constraint, so that reliability is necessary but not sufficient for justification.  The subject’s perspective, or matters internal to the subject, may be introduced in various ways. A person’s having reasons to believe that some process is unreliable or that some ground is inadequate may constitute overriding factors for a belief’s prima facie justification. Ultima facie or unqualified justification will depend on there not being any significant overriders for the belief (alternatively, the belief’s being supported by the totality of the person’s knowledge, belief, and experience). Also we may state that a belief, to be justified, must be based on something, a ground, which is internal to the subject in that it constitutes a psychological state of the person in the form of a belief or an experience. But the ground’s adequacy or justificatory efficacy amounts to the ground being a reliable indicator of the truth of the belief. Alston developments the internal constraints suggested here in his reliable indication theory of justification, according to which a belief is justified if and only if it is based upon an adequate ground. And an adequate ground will be a ground that is  “sufficiently indicative of the truth of p.” Alston claims that: “. . .a belief’s being justified is a favorable status vis-à-vis the basic aim of believing, or, more generally, of cognition, viz., to believe truly rather than falsely.  For a ground to be favorable relative to this aim it must be ‘truth conducive;’ it must be sufficiently indicative of the truth of the belief it grounds.  In other terms, the ground must be such that the probability of the belief’s being true, given that ground, is very high” (1989c, p. 232).

 

D. Probability and Evidence

 

            The notion of probability frequently enters into reliabilist accounts to explicate the notion of truth-conducivity. The kind of probability Alston has in mind is objective probability, though the kind of objective probability he does not spell out with any detail, holding that in fact no adequate concept of it has been developed. He says that he is thinking in terms of “some kind of tendency” related to “the lawful structure of the world” whereby “one state of affairs renders another probable” (1989c, p. 232). Typical of reliabilism, Alston has in view physical or statistical probabilities (see footnote no. 5). But in his “Concepts of Epistemic Justification” Alston suggests that the model of conditional probability be used to explicate the notion of truth-conducive adequacy.  A ground G will be adequate just if the probability of p on G is high. Of course, a necessary truth will have an objective probability of 1 on any evidence whatever. The formula implies that a person’s belief that, say <everything which is red is coloured> will be based on an adequate ground even if he believes it on the basis of it being asserted as true by his favourite character in a television cartoon. Any bit of evidence will function as an adequate ground for a necessary truth since such truths have a maximal probability on any bit of evidence. Conversely, necessary falsehoods could never have an adequate ground since they have an objective probability of zero on any evidence, so one could never be truth-conducively justified in holding such propositions.  The idea of “objective probability,” then, is slippery and needs further refinement.[29]

            Marshall Swain’s account of a probabilistic-reliability model of epistemic justification (Swain 1981, pp. 96-120) is perhaps best suited to tighten up Alston’s reliable indication account. According to Swain, ascriptions of reliability are evidential claims that may be stated in terms of conditional inductive probability: “S’s believing that h on the basis of R is epistemically justified at t iff: S’s believing that h on the basis of R is a reliable indication that h is true” (1981, p. 99). Given an evidential base E, where E = [(1) the belief that h on the basis of R and (2) a set of the believer’s relevant characteristics], when the probability of h on E is greater than the probability of not-h given E we have a reliable indication of the way the world is and so an adequate ground for the belief that h.  Swain wants to take into account not only the reasons a person has for his belief h (which must themselves be justified) but also the characteristics a person has which are relevant to the reliable gathering and processing of information.[30]  Such characteristics must, of course, neither entail h nor not-h (excepting cases where h is a necessary truth), and they must be probabilistically neutral with respect to h when taken alone (so that the characteristics affect conditional probabilities only when taken in conjunction with (1)).  Now if we suppose that h is a necessary truth (and so has an objective probability of 1 on any evidence), what will preclude a person S from being justified in believing h is S’s lacking some reliability-relevant characteristic or possessing some relevant unreliability characteristic.  If S believes that <everything that is red is coloured> because he heard it from a cartoon character, and S has a habit of believing as true what cartoon characters say is true, then although the proposition S believes is true (and necessarily so) on the evidence in question, it is nonetheless accidental that S believes a true proposition given that same evidence. The distinction here (also made by Swain 1981, pp. 111-112) is between the probability of a necessary proposition Np on the basis of some evidence e and the probability that <S believes some necessarily true proposition Np on the basis of R>. Let this latter bracketed proposition by symbolized by h*, and let e = (i) S’s (unreliable) characteristics and (ii) the further proposition q*: <S believes some proposition or other on the basis of R>. Even if e is good evidence for Np, it is not good evidence for h*. Although the probability of Np will be 1 given e, the probability of h* given e will not be greater than the probability of not-h* given e.

            To locate the Swainian account of conditional probability within the larger corpus of concepts of evidential probability it is important to note the different senses in which we can speak of the probability of some hypothesis h on contingent evidence e. Assuming that there are right and wrong ways to assess the relevance and force of evidence, calculating the value of h on e will require the application of correct inductive standards. These criteria furnish us with necessary truths about what is evidence for what and how strong particular evidence is. They are essential to our coming to correct judgements about the force of h on e. If, per impossibile, human cognizers knew every logical truth (including the deductive consequences of all contingent evidence and the relevant possibilities and necessities), and all values of prior probabilities on tautological evidence, they would know the total force of h on e. Such would be the value of h for any logically omniscient being. His judgement would be equivalent to the logical probability of h on e. His calculation of P(h/e) would include every logical truth in the “evidence” slot, for (i) any bit of contingent evidence entails all logical truths and (ii) being unlimited in his logical capacities he would know all the deductive consequences of e. Consequently, the value of h on (any) e could never be altered since no logical truth could ever be discovered which wasn’t already factored into the calculation.

            Logical probability (a kind of objective probability) may be contrasted with epistemic evidential probability, which involves the calculation of the value of h on e by human cognizers with limited logical capacities and knowledge but nonetheless correct inductive standards. Although a less than logically omniscient being may hold no false logical beliefs (and no invalid rules of inference), he may be ignorant of any number of deductive consequences of his evidence, as well as what possibilities and necessities are relevant for determining the value of h (both of which are determinants of the true value of h). So he will not know the total force of h on e, and the calculated value of h on e at t₁ will be affected (i.e., increased or decreased) by new observational evidence and the discovery of new logical truths at t_n+1.[31]  Another kind of evidential probability arises when the value of h on e is calculated relative to our contingent evidence, logical capacity, and different inductive and deductive criteria. Logical and epistemic probabilities assume that P(h/e) is calculated by the application of correct inductive standards, but evidence may have a certain force for a person S given S’s own inductive standards - how S views the force of the evidence. Such a probability we can call subjective evidential probability. [32]

            The Alstonian-Swainian perspective on probability conditionalizes on an evidence-base which includes certain reliability-characteristics of S. In this way it goes beyond the evidence S has for p - what S is aware of. Part of this seems to follow from the reliabilist focus on beliefs as opposed to propositions. As pointed out above, what fills the “h” slot is not the proposition p (the potential object of belief) but S’s believing a true proposition given the evidence he has for p and contingent facts about S’s epistemically relevant characteristics.[33]   A nonreliabilist approach will develop the idea that what is required for justification is that the proposition p (which S believes) is rendered probable by correct inductive standards (or one’s own inductive standards in the sense of subjective probability) which can be grasped a priori. Whatever psychological interest there is in the causal source of a belief, the conditional probability of h on e will be the probability of h on the evidence one has in support of p. The extent to which e renders h probable (given the application of some inductive criteria) to that extent one has indications that p is true (even if one knows nothing of the contingent empirical features of one’s processes of belief formation).

 

III. Internalism and Externalism in Epistemology

 

            The preceding brings us to yet another way of characterizing the pluralism in contemporary epistemology: distinguishing between those accounts of justification (or knowledge) which are internalist and those which are externalist.  The deontological concept of justification has frequently been associated with internalism; reliabilism with externalism.  Internalist theories of justification (or knowledge) maintain that what confers justification (or alternatively, what transforms true belief into knowledge) is restricted to items within the believing subject. These are either something within the believer's perspective of the world (i.e., other beliefs or knowledge) or as something (in some sense and to some degree) cognitively accessible upon reflection.  Externalism, by contrast, is simply the denial that (at least some of) these internalist constraints hold. Justification is conferred by factors or properties to which the cognizer may have no special cognitive access or no epistemic access at all.

 

A. Internalist Theories of Justification

 

            There is a long tradition in the history of Western epistemology from Descartes onward that views knowledge as the result of reflective inquiry. In fact, one might say that the dominant viewpoint has been this tradition of epistemic reflectivism according to which knowledge requires a cognitive grasp or awareness of the conditions which warrant a given belief, whether that warrant be established by reasons, evidence, or something of the sort. In other terms, the Western tradition (at least since Descartes) has been dominated by an “internalist” viewpoint.  Where knowledge has been understood as “justified true belief,” the notion of justification has been governed by internalist constraints. As stated above, internalism requires that the subject have, in some form or another, a cognitive grasp of the justifiers or the efficacy of the justifying conditions of belief.  What makes a belief justified, or what transforms true belief into knowledge, is “within” the individual.

            Internalism, though, comes in several forms.  There is first what has been called perspectival internalism (hereafter PI), according to which justifiers are restricted to what is within the cognizer’s viewpoint, and where this “viewpoint” is constituted by the cognizer’s knowledge (strong PI), justified beliefs (moderate PI), or simply beliefs (weak PI).  The basic principle here, of course, is that of a doxastic state:  the only thing that can justify some person S’s belief that p is some other doxastic state. The strength of the particular version of PI will depend on the epistemic status of the propositional attitude. A more complex version of internalism is access internalism (hereafter AI), according to which a justifier must be the sort of thing which is cognitively accessible to the subject,  something of which the subject  is capable (perhaps immediately so) of being aware of upon reflection.  This would include experiences (broadly construed as inclusive of sensations, feelings, and a proposition’s seeming obvious to one) as well as beliefs. So this constitutes a kind of broadening of PI by including what is potentially in the subject’s viewpoint, since those items are what the subject could come to know (or justifiably believe) upon reflection.[34]   Varieties of AI are easily generated by chisholming the “accessibility” condition in terms of both the kind and degree of accessibility. A strong kind of accessibility would be the requirement that the justifier be accessible to the actual subject’s consciousness. A more moderate version would require only that it be the sort of thing which is accessible to people in general, typically accessible to normal subjects.  In addition, we can establish conditions regarding “the degree” to which a person has access to the ground of some belief that p.  Some (Ginet 1975) speak of “directly recognizable;” others “fairly direct accessibility” (Alston 1989c, p. 238).[35]

            Discussions of internalism frequently overlook or otherwise obscure another important distinction, whether the appropriate internalist constraints (PI or AI) are imposed on the grounds of belief or the adequacy of the grounds. An internalist with respect to only the grounds of some belief that p maintains that a person is justified in believing p only if the ground of p is "within" the subject. The ground must be other justified beliefs (or knowledge) of the subject (PI) or something which the subject can be aware of by turning his attention to it (AI), experiences or beliefs, even if the justificatory efficacy of the ground is not accessible. We can think of this as first-order or lower level PI and AI.

 

PI¹       S is justified in believing that p only if S’s belief that p is based upon other justified beliefs (or knowledge) of S.

 

AI¹       S is justified in believing that p only if S’s belief that p is based on a ground which is accessible to S fairly readily on the basis of reflection.

 

It might be thought, though, that the content of at least one of the beliefs which constitutes the ground of S’s belief that p is a belief about the adequacy of the ground of the belief that p, or that adequacy in some sense of AI is cognitvely accessible. This gives rise to second-order internalism.[36]

 

PI²       S is justified in believing that p only if S’s belief that p is based on S’s justified belief that <the ground of the belief that p is an adequate one>.

 

AI²       S is justified in believing that p only if S is capable, fairly readily on the basis of reflection, to acquire a justified belief that <the ground of S’s belief that p is an adequate one>.

 

 

            AI’s internalist constraint requires that a ground be internal in the sense that the subject has cognitive access to it, as either a belief or an experience. When the internalist constraint of AI is imposed upon adequacy, the question becomes one of “accessibility” to the adequacy of a ground.  Alston suggests that this accessibility might be the capacity of the subject to come into the state required by PI². AI² differs from PI² in that it does not require the possession of an actual justified higher-level belief in the adequacy of the grounds. Rather, AI² requires that for each justified belief that p, it is possible for S to acquire upon reflection a higher-level belief to the effect that the grounds are adequate. So where the ground is some experience e, S is justified in believing that p only if S is capable upon reflection of (justifiably) believing that e is an adequate indication that p. AI² could be weakened by altering the nature of the accessibility requirement. One could maintain that the conditions stipulated in AI² be true only “in general” not in every case. The adequacy of the ground would only have to be the sort of thing that is typically accessible fairly readily upon reflection.[37]

 

B.  Critical Assessment of Internalism

 

            Upon closer scrutiny, some versions of internalism turn out to be radically implausible. PI¹ and PI² are two such cases. PI¹ entails either an infinite regress or some form of coherentism (linear or holistic), both of which I have argued are dubious indeed.  In both cases, internalism would leave us with no immediately justified beliefs (or at least no wholly immediate justification), and that simply seems absurd. Perceptual beliefs and beliefs about one’s own current conscious states do not require being based upon other justified beliefs. I can, for instance, be justified in believing that “I feel tired,” “I see Professor Ward walking down Cornmarket,” or “I hear an Elvis Presley song of the radio” without recourse, consciously or unconsciously, to propositional attitudes which make up my view of the world.  Much less does it seem plausible to suppose that I am justified in such perceptual or introspective beliefs only if I have or base them in part on beliefs about the adequacy of grounds or the reliability of belief-forming processes. Many subjects do not have adequacy beliefs and yet we affirm justification for them (in a truth-conducive sense). Moreover, PI² yields an infinite regress similar to PI¹. If S must justifiably believe that the reasons of p are adequate in order to be justified in believing that p, then the belief that the reason adequately supports p, call it r, is itself subject to a higher-level requirement.  And so, if S’s initial belief that p is justified, then an infinite hierarchy of beliefs must be justified in an ascent of levels. Unlike PI¹ there is no coherentist alternative to infinite regress here just because the regress would involve an infinite doxastic hierarchy in which at each level there is a belief about adequacy relations holding at a previous level. A circular chain would require doubling back at some point to an adequacy-belief at a previous stage which would then be required to justify an adequacy-belief at a later stage. The difference in content between adequacy-beliefs at different levels precludes such circular justification.[38]

            Two other points are worth noting in relation to PI².  Even more so than PI¹, PI² seems to be a consequence of conflating the activity of justifying a belief with the state of being justified. If one is to show that one is justified, one must adduce reasons for regarding one’s belief as justified, and in that case one will need recourse to beliefs about the efficacy of the justifiers, the adequacy or reliability of the grounds, etc. Earlier I claimed that showing justification is not necessary to the state of being justified. Also, PI² shows the connection between internalism and deontologism with greater lucidity than PI¹ . Recalling the account of deontologism, the sorts of intellectual obligations that typically fall on a subject relate to the epistemic point of view. Even if deontological justification does not entail truth-conducive justification, our most important intellectual obligations (if there are such things) will require our conducting our cognitive life in such a way that as far as we can see we are maximizing truth and minimizing falsity in our believings. But this would seem to require the adequacy of support to be cognitively accessible. My merely believing that p on the basis of q does not constitute intellectual dutifulness, unless I am justified in supposing that q provides the appropriate sort of support for p.  Otherwise I would not be conducting my intellectual life in such a way that it aligns itself to the alethic goal of believing. Under a deontological conception a person is properly held to be epistemically irresponsible if he forms beliefs on the basis of reasons that he does not have good reasons to believe are adequate and so contribute towards the epistemic point of view.[39]

            In contrast to the versions of PI, AI² is only moderately implausible. AI² does not appear to suffer from the epistemically exaggerated condition of an actual infinite hierarchy of justified beliefs in the support relations.  The most that can be said of the infinite regress on an AI² construal is that it is only a potential regress. Nor does it restrict justifiers to other beliefs, and so leaves open the possibility of immediate justification. Nevertheless, AI² seems to be too strong of an accessibility demand for justification. Is it plausible to hold that S is justified in the belief that p only if S possesses the capacity to come to a higher-level belief that the grounds are adequate?  I think not for the majority of epistemic subjects.  This is obviously true for any construal of accessibility that requires “directly recognizable” adequacy relations.  Our noetic structures are complex structures, and it is excessively optimistic to hold that we can simply come justifiably to believe that grounds are adequate just on reflection at the moment. Sometimes the justifying relations might involve a large number of propositions, more than the normal person could entertain at one time, and certainly not immediately upon reflection. We are accustomed to perceptual cues often being subtle, but this difficulty of immediately discerning or specifying what a belief is based upon applies a fortiori to the adequacy of grounds. The complexity of support relations often makes direct access to them difficult, if not impossible. Even weaker degrees of accessibility make AI² worrisome. For one thing, AI² entails a certain level of conceptual sophistication that is absent in many epistemic subjects (e.g., little children). The whole question of the adequacy of the grounds of belief never enters into the mind of many people who, nevertheless, form justified beliefs over a whole range of subjects. One would be hard-pressed to imagine a child even having the capacity for raising the question of the adequacy of the grounds for her belief that there are three dolls in her doll house or that 3 + 2 = 5. And it seems that not a few adults have had problems with having access to the support relations of certain beliefs.  Moreover, where “adequacy” must be the sort of thing that is cognitively accessible upon reflection, certain constraints fall on what adequacy could amount to. Obviously if adequacy is spelled out in terms of contingent features of our epistemic situation (e.g., reliable processes of belief formation or statistical or propensity theories of probability), as most truth-conducive accounts of justification maintain, this kind of adequacy is not accessible just upon reflection, but would require some sort of empirical investigation. So, for instance, whether a given sensory perceptual belief is based upon an adequate ground will in part be a function of the reliability of the perceptual belief-forming mechanisms involved. Our determining whether such mechanisms are (and were in any particular case) reliable depends on what we (allegedly) know about our sensory perceptual mechanisms and environment from psychology, physiology, and the natural and physical sciences. Consequently, on AI² what is accessible must be certain beliefs about the adequacy of grounds or reliability of processes (which is obviously compatible with such grounds not in fact being adequate indications of the truth of p).

            The above considerations suggest that most forms of internalism are highly untenable. PI on both levels entails an infinite regress, and thereby prevents the subject from ever being justified in a single belief. AI constraints on the higher level, even in their weaker forms, though not suffering from the regress dilemma, do share with PI constraints unrealistic demands for the level or degree of cognitive awareness or access to the adequacy of the grounds of belief. And both PI² and AI² might stem from deontologism and a conflation of being justified with showing justification, and thereby confuse structural and dialectical issues in epistemology. Thus we are left only with AI¹ as a plausible internalist constraint.

 

C.  Externalism and Mixed Theories

 

            Epistemic externalism, on the other hand, is a denial of internalist constraints, and since internalist constraints may be spelled out in a variety of ways, so also with externalism. Typically, though, the externalist will hold that one need not have any cognitive awareness of the justifying conditions of one’s (justified) beliefs. The thing that confers justification is, in some way, external to the individual.[40]  One of the attractions of externalism is that some philosophers consider it to be a solution to the epistemic regress problem. The foundationalist has to face the question as to how basic beliefs can confer justification to nonbasic beliefs without themselves being mediately justified.  Basic beliefs must be justified if they are to transfer a justificatory status to some other belief(s), but if this justification rests upon other beliefs (as in PI) then we are cast into the nasty regress problem discussed earlier. Let us suppose, though, that basic beliefs are justified not by virtue of their relation to other justified beliefs of the subject. Rather they are justified by some external state of affairs, such that these conditions render the belief likely to be true, then we can avoid the regress and yet preserve the element of truth-conducivity essential to knowledge and many theories of epistemic justification.  Externalism also involves a source-relevant position on justification (encountered earlier in the Alstonian-Swainian conditionalization of reliability), and so sits well with the intuition that the causal origin or what is causally operative in sustaining a belief is essential to its justificational or epistemic status. Suppose Julie has good evidence (in the form of other beliefs or knowledge) that her next door neighbor Sam is stealing her Sunday morning newspaper. She also believes this. But suppose that she believes it, not because of the evidence she has for it, but because of paranoia and the fact that she dreamed she saw him stealing her paper. It would strike many that, under such circumstances, Julie’s belief is not justified or does not constitute knowledge. This tendency is plausibly explained by the sense that what is causally responsible for a person’s holding a belief is essential to its epistemic status.[41]  Although the source-relevant view of justification is compatible with internalism, it is typically associated with and emphasized by externalism. Moreover, we have seen in Alston and Swain how the source-relevant view is combined with a fairly strong position on truth-conducivity (not typical for internalism).  Externalism may be thought to offer the most straightforward way of satisfying the epistemic point of view since its conception of “likelihood of truth” is identified with objective probabilities which yield actual truth-conducivity rather than (justified) beliefs about the truth-conducivity of our believings.

            Notwithstanding these virtues, the main pitfalls appear to be encountered in the necessarily perspectival character of justification. I noted earlier that the inclination to identify justification with reliability because of the epistemic point of view is circumvented by internalist intuitions regarding “justification.” A person cannot be justified in some belief merely because of external relations that hold between that belief and some state of affairs independent of the subject’s perspective. Mere truth-conducivity is not sufficient. Externalism seems to run counter to intuitions that justification is in some sense perspectival, involves the subject’s own evaluations and perspective. Also, since "truth indications" are not accessible according to the externalist, it would seem that we are left without a theory of justification that offers us a way to sift through our beliefs and best guide us in acquiring true beliefs (Bonjour 1980, p. 63). What may be essential to justification, then, is not merely that beliefs are generated in a reliable way or that grounds are a reliable indication of the truth of the belief but that we have reason to regard a belief as likely to be true. Some of these criticisms of externalism seem to stem from epistemic level confusions (Alston 1989c, pp. 112-114, 153-171). I have already noted that higher-level requirements for justification (such as the belief that “reason R is an adequate reason for p” or “experience E is an adequate indication that p”) are unnecessary and improper requirements for epistemic justification (especially where actual truth-conducivity is in view). What such higher-level requirements seem to be required for is being epistemically justified in the higher-level belief that one is epistemically justified. But one can be justified simpliciter without believing that one is justified and one can believe that one is justified without in fact being justified.[42]  Of course, there may be an important distinction between conditions necessary for being epistemically justified and requirements for being a responsible truth-seeker (where we are concerned with regulative matters for our doxastic life). And it may be that higher-level justification plays a role in the latter even if it is not central to the former. Nonetheless, there is merit in the contention that justification requires some kind of internalist component.

            It need not be thought though that internalism and externalism constitute rigid boundaries with no middle ground. The diversity of ways of spelling each out precludes this, but it also provides a way of bringing both elements together. Source-relevance is easily introduced by making what a person’s belief is based on, rather than what evidence they have, as necessary to justification. Distinguishing between grounds and their adequacy allows further possibilities.  One might hold to a PI¹ requirement on grounds and adopt externalism with respect to the adequacy of those grounds. To allow for immediate justification (and perhaps avoid infinite regress problems), a more plausible line would be to take an AI¹ position on grounds and maintain an externalist position on the adequacy of the grounds. This latter suggestion we have seen in relation to Alston’s account of justification as a belief’s being based upon an (internally accessible) but (objectively) adequate ground, where the ground consists of a person’s other psychological states: experiences and other beliefs, or knowledge. But the distinction between prima facie and ultima facie (or all things considered) justification may also be applied to blend externalism and internalism. A prima facie justification is one that can be nullified by certain overriding reasons. Such reasons may be sufficient for believing that p is false (rebutters), or they may be reasons such that the conjunction of them and the ground of p entails that the ground of p is not sufficiently indicative of the truth of p (undercutters).  Note, however, that overriding reasons are “reasons” and therefore fall into the PI category. An internalist might combine AI¹ and a restricted PI¹ (or PI²) then, by maintaining that AI¹ constitutes a requirement for prima facie justification, but that this justification can be overridden by the constraints imposed by a negative PI¹ (or PI²) constraint. Externalism could then be adopted again with reference to the adequacy of the grounds. There is no reason to hold the radical externalist position that a person’s belief is justified even if they have good reasons for regarding it as unjustified.

            There is much to be said for the internalist ramifications of defeaters for justification, especially when we work into the picture the distinction between belief-formation and sustenance. For instance, it may be that if one has a defeater for a belief, one only remains justified in that belief if one has a defeater for this initial defeater (a so-called defeater-defeater). And although a defeater-defeater need not be another belief-state, it clearly could be; indeed, for some people at certain times, and under certain conditions, it might have to be. One could argue that for some people on particular occasions, given a PI defeater, their remaining justified in believing that p requires that they have a PI defeater-defeater (i.e., a reason to believe that their belief is true or that its grounds are adequate). Further permutations are possible by distinguishing between evidential aspects of justification and the psychological source for a belief. I shall develop such prospects in the course of the thesis. The conclusion for the moment is that even if one’s concept of justification is highly externalist, it may also have an internalist twist. And however externalist one is regarding the concept of justification the conditions for justification of a belief or “this” or “that” sort may even require further internalist constraints.

 

IV.  Conclusion

 

            In this chapter I have aimed at a critical exposition of crucial issues in contemporary epistemology. Two themes have dominated: (1) the plausibility and prima facie advantage of modest foundationalism over classical foundationalism and various forms of coherentism and (2) the positive prospects for combining internalism and externalism. In the course of the thesis both themes will be played out in further detail with respect to theistic belief in particular. My central thesis is that there is an epistemically adequate form of evidentialism that is compatible with the religious epistemology of Alvin Plantinga. Foundationalism and the confrontation between internalism and externalism loom large in the discussion. From my thesis will emerge a version of theistic foundationalism which is at once compatible with the strong externalism of Reformed epistemology but which satisfies what I will argue are the important internalist intuitions behind classical evidentialism.