Skepticism and Naturalistic Epistemology



In this essay I am primarily interested in exploring the implications of contemporary versions of naturalistic epistemology for the way in which one should address traditional skeptical arguments. My interest goes beyond this, however, for I remain convinced that once one understands clearly the full implications of such views, one should find their approach to philosophical epistemology implausible.




Before we proceed, however, a few comments about the concept of naturalism as it relates to such fields as epistemology, philosophy of mind, ethics, mathematics, and the like, might be in order. A common rhetorical technique of philosophers involves the attempt to seize a terminological high ground. Thus, the philosopher who succeeds in getting an opponent's theory of perception called "naive realism" has the battle half won. And if one's philosophical opponents manage to get their views labeled "naturalistic" it might appear that one is left with the unenviable task of defending "unnatural" or "nonnatural" views. But it is obviously important right from the start to remember that we are dealing with technical philosophical debates framed in technical philosophical vocabulary. And it is more than a little difficult to see what is so "natural" about naturalistic philosophical views.


What does make a view naturalistic? What is the contrast that is intended to be drawn? Generally speaking, the naturalists are on the side of what they take to be scientifically respectable view. Naturalists typically analyze the properties with which they are interested into natural properties. And what makes a property natural? That it is the kind of property with which science feels comfortable. Of course, the concepts which scientists take to be unproblematic vary depending on the nature of the science, and thus "respectable" naturalism is relative to the science with which the naturalistic philosopher identifies. Thus historically prominent naturalists in ethics include among the natural properties such psychological properties as being desired. But the philosopher of mind who wants a naturalistic account of mental states might well reject the "folk science" vocabulary of desire as unsuited to a philosophically perspicuous reduction of mental states. That philosopher might feel more comfortable with an analysans that employs the concepts of contemporary cognitive science.


Why do so many contemporary philosophers want philosophical analyses provided in the vocabulary of science? The most obvious answer, I suppose, is that science has enjoyed extraordinary success. Television sets, computers, and microwave ovens convince many that whatever they are doing, scientists are doing it effectively. The terms they employ in their scientific reasoning must unproblematically refer to kinds of things that really exist. If we as philosophers want to piggy-back on the success of science then we ought to make sure that we can reduce traditional philosophical disputes to terms that make the relevant questions scientifically respectable. Typically, the "harder" the science, the more impressive has been its successes, and consequently if we can reduce familiar concepts of mental states to the concepts of physics, we will be better off than if we stay with the concepts employed by the relatively soft science of psychology. Psychologists, after all, are hopeless when it comes to offering the kind of explanations that offer the possibility of lawful prediction. Indeed, they are hopeless when it comes to offering anything that looks like a genuine law of nature.


There is, of course, a real irony in the fact that so many philosophers yearn to provide philosophical analyses employing the "respectable" concepts of contemporary science. One of the concepts most eagerly embraced by naturalists for use in naturalistic analyses is the concept of causal and lawful connection. Thus many naturalists seek to understand intentionality and representational capacity in terms of complex causal connections that obtain between the thing represented and the entity or event doing the representing.1 Many naturalistic epistemologists seek to understand knowledge and justified belief by understanding the relation that exists between knower and known as like that obtaining between thermometer and the temperature it registers.2 Others embrace the view that justified belief is essentially reliably produced belief where the concept of reliability is clearly to be understood in terms of lawful (probabilistic or universal) connections between the conditions that produce the belief and the truth of the belief produced. But while the concept of causal and lawful connection does seem to be presupposed by scientific inquiry, it is, from a certain perspective, one of the least respectable concepts for use in philosophical analyses.


The history of philosophical attempts to provide illuminating analyses of the concept of lawful connection, causal connection, or the relation expressed by contingent subjunctive conditionals is hardly a history distinguished by its glowing success. It would be a gross understatement to suggest that there is no philosophically received view about the correct way to understand any of these unproblematic concepts. Many still embrace Humean regularity theories as the most promising attempt to get at the heart of causal or lawful connection, but no one has offered a plausible account of the distinction between lawful and accidental regularity within the framework of a Humean analysis. Recently, a number of philosophers have suggested that we think of lawful connection as a kind of sui generis relation obtaining between universals,3 but again, it is a distinct minority of philosophers who can convince themselves that they are either phenomenologically acquainted with such a relation or have dialectically defensible reasons for positing its existence. Possible worlds accounts of lawful connection, causal connection, or the truth conditions of counterfactuals are almost comical. The interpretation of the relevant possible worlds semantics inevitably requires an understanding of the very modalities one seeks to illuminate.


With the embarrassing lack of success that philosophers have had arriving at enlightening accounts of lawful and causal connection, it is a little odd to find so many philosophers scrambling to make philosophical accounts of intentionality and epistemic concepts "respectable" by reducing the relevant intentional and epistemic properties to nomological properties. And it is more than a little bit hard to listen to such philosophers bemoaning the philosopher of mind who appeals to a "mysterious" sui generis capacity of mental properties to constitute "natural" signs of the states of affairs they represent, or the equally "mysterious" relations of acquaintance to facts that is said by some philosophers to partially constitute the foundations for empirical knowledge.


It is worth remembering that given an epistemic perspective that was virtually taken for granted for hundreds, if not thousands, of years, it is facts about mental states that have been taken to be conceptually and epistemically more secure than any of the complex facts that form the subject matter of contemporary science. The traditional problem of perception, after all, just is the problem of getting philosophically respectable access to the physical world with which science begins its investigation. In the heyday of logical positivism at least one goal was to reduce scientific talk into an analytically equivalent vocabulary of sense data, precisely because sense data were taken to be conceptually and epistemically less problematic than physical objects. Furthermore, even if these philosophical theories about what is and is not conceptually and epistemically unproblematic are not defensible, it is worth remembering that the sciences studying the mind can get nowhere without presupposing in the course of their investigations that people have unproblematic access to their mental life. If people do not know when they are in pain, for example, the cognitive psychologist will have to wait until the cows come home to make any discoveries about the state of the brain responsible for pain.


Now it is obviously unprofitable to continue at length this rather abstract (and somewhat polemical) discussion of the relative merits of naturalist as opposed to nonnaturalist accounts of philosophically interesting concepts. It is, however, worth emphasizing one of the initial motivations behind naturalistic epistemology that goes beyond any of the factors mentioned above. Specifically, the naturalistic epistemologist seems moved, in large part, by the conviction that it is only by taking a naturalistic turn that the epistemologist can avoid massive skepticism. Quine's famous injunction to naturalize epistemology seemed largely based on the alleged impossibility of refuting the skeptical challenge from within traditional foundationalist frameworks.4 In the remainder of this essay I want to examine more precisely how naturalizing epistemology in the manner of contemporary externalists would affect the way in which we respond to traditional skeptical arguments. Again, as I see it, contemporary externalists are naturalistic epistemologists because of their commitment to analyzing fundamental epistemic concepts into "natural" nomological properties and because of the fact that their views allow them to follow Quine's advice to use the methods of natural science to answer epistemological questions (more about this later).5 For much of the discussion I will take as my paradigm naturalistic epistemology the kind of reliabilism that Goldman first defended in Goldman (1979). What I say will apply mutatis mutandis to other paradigm naturalistic/externalist accounts of epistemic concepts such as Armstrong's conception of basic knowledge (1973) and Nozick's much discussed "tracking" analysis of knowledge and evidence (Nozick, 1981).




Let me very briefly sketch the kind of skepticism with which I am concerned and what I take to be the classic form of the skeptical challenge. First, it might be helpful to emphasize that the problem of skepticism is primarily a problem in what I call normative epistemology. The contrast is to metaepistemology. Metaepistemological questions concern the philosophical analysis of fundamental epistemic concepts. Normative epistemological questions ask to what the epistemic concepts apply.6 As I understand it, naturalism in epistemology is best construed as a metaepistemological position.7 Claims that we are or are not justified in believing something are part of normative epistemology.


The skepticism I am most interested in is skepticism with respect to justified or epistemically rational belief (rather than knowledge). Furthermore, I am primarily concerned with what we might call "local" rather than "global" skepticism. The global skeptic maintains the highly suspicious (because it is epistemically self-refuting 8) position that we have no justification for believing anything. The local skeptic puts forth arguments designed to establish that we have no justified beliefs with respect to a certain class of propositions. The history of local skepticisms includes skeptics with respect to the possibility of justifying beliefs about the physical world, the past, other minds, the future, and so on. The traditional skeptic virtually always presupposed some version of foundationalism, i.e., presupposed that we do have noninferentially justified belief in at least some propositions. To be sure the presupposition was seldom explicitly stated, but one cannot read any of the important historical figures concerned with either advancing or refuting skepticism without reaching the conclusion that they took some propositions to be epistemically unproblematic, where their unproblematic character seemed to stem from the fact that one did not need to infer their truth from any other propositions believed.9


The first step, then. in advancing an argument for skepticism with respect to some kind of proposition is to establish that our access to the relevant truth is at best indirect. Skeptics with respect to the physical world deny that we have noninferential "direct" access to physical objects. The standard skeptical claim is that if we have justification for believing anything about the physical world, that justification reduces to what we can legitimately infer about the physical world from what we know about the character of our past and present sensations. The skeptic about the past claims that we have no direct (i.e.. noninferential) access to truths about the past. What we know or reasonably believe about the past is restricted to what we can legitimately infer about past events from what we know about the present state of our minds. The future is known to us only through our knowledge of the past and present. Other minds are known to us only through what we know about the physical appearance and behavior of bodies.





As many of its proponents would be quick to point out, paradigmatic naturalistic (and externalist) accounts of justification actually allow a distinction that parallels the classic distinction between noninferentially and inferentially justified beliefs, The reliabilist, for example, can distinguish reliable belief- producing processes that take as their input belief states from reliable belief- producing processes that take as their input something other than belief states. And it would be natural to describe justified beliefs produced by the latter as noninferentially justified beliefs. Their justification, after all, would not depend on the having of any other different justified beliefs. Belief-dependent processes can generate justified beliefs only when they take as input beliefs that are already justified.10


Now one of the primary advantages that paradigmatic naturalistic accounts might seem to have in the battle against skepticism is the ease with which they can deny the crucial first premise of skeptical arguments. The class of noninferentially justified beliefs is likely to be much larger given a reliabilist or "tracking" account of justified belief. Notice, I say "likely" to be much larger. As far as I can see, virtually all naturalistic epistemologies entail that it is a purely contingent question as to which beliefs are justified noninferentially and which are not. On the reliabilist's view, for example. the question of whether or not one is noninferentially justified in believing at least some propositions about the physical world is a question about the nature of the processes that yield beliefs about the physical world and the nature of their "input." If we have been programmed through evolution to react to sensory stimuli with certain representations of the world, and we have been lucky enough to have "effective" programming, then we will have noninferentially justified beliefs about the physical world. If Nozick (1981) is right and our beliefs track facts about the physical world around us, and this tracking does not involve inference from other propositions, we will again have noninferentially justified beliefs about the physical world. If our beliefs about the physical world are acting like that reliable thermometer that Armstrong (1973) uses as his model for direct knowledge, if we are accurately registering the physical world around us with the appropriate representations, then again we have noninferential, direct knowledge of the world. Whether or not we have such noninferential justification for believing propositions describing the physical world, on any of these naturalistic ways of understanding noninferential justification, is a purely contingent matter.


Now that it is a contingent fact is not in itself surprising, nor is it a consequence peculiar to naturalistic epistemology. It is certainly a contingent fact on a classic "acquaintance" foundationalist theory'' that I am acquainted with the fact that I am in pain. It is a contingent fact that I am in pain and so obviously contingent that I am acquainted with it. It is less obvious on traditional foundationalisms that it is a contingent fact that we are not acquainted with certain facts. It might seem, for example, that one could not be acquainted with facts about the distant past, the future, or even the physical world if it is understood as a construct out of actual or possible experience or as the cause of certain actual and possible experience.12 But even here it is difficult to claim that it is necessarily the case that conscious minds are not acquainted with such facts. There may be no God but it is not obvious that the concept of a consciousness far greater than ours is unintelligible. If the concept of a specious present makes sense, such a consciousness may have the capacity to directly apprehend a much greater expanse of time than can finite minds, In any event it is not clear that the class of facts with which we can be acquainted exhausts the facts with which all possible consciousness can be acquainted.


But even if the scope of noninferentially justified belief is contingent on both classic foundationalist views and contemporary naturalistic (externalist) versions of foundationalism, there are crucial differences. On traditional versions of foundationalism, philosophers are at least in a position to address reasonably the question of the content of noninferentially justified belief. The philosopher is competent, at least as competent as anyone else, to address the questions of whether or not we have noninferentially justified beliefs in propositions about the physical world, for example. There are two sources of knowledge as to what we are noninferentially justified in believing. One is dialectical argument. The other is acquaintance itself. One can be directly acquainted with the fact that one is directly acquainted with certain facts.


According to paradigm naturalistic epistemology, the facts that determine whether one is noninferentially justified in believing a proposition are complex nomological facts. Given such views, it is not clear that a philosopher qua philosopher is even in a position to speculate intelligently on the question of whether or not we have noninferentially justified belief in any of the propositions under skeptical attack.13 Because the naturalist has reduced the question of what is noninferentially justified to questions about the nature of the causal interaction between stimuli and response. and particularly to the processes of the brain that operate on the stimuli so as to produce the response, the search for noninferential justification would seem to be as much in the purview of the neurophysiologist as the philosopher.14 In the last two hundred years, the vast majority of philosophers simply have not had the training to do a decent job of investigating the hardware and software of the brain. But without this training, it hardly seems reasonable for philosophers to be speculating as to what is or is not a reliable belief-independent process. To be sure, some contemporary epistemologists are trying to "catch-up" with developments in cognitive science and even neurophysiology, but I cannot help worrying that the experts in such fields will quite correctly regard these philosophers as simply dilettantes who, having tired of their a priori discipline, now want to get their hands dirty in the real-life world of science.


Given the above, it is ironic that so many philosophers find naturalistic epistemology attractive precisely because it can more easily capture the prephilosophical intuition that there is something direct about our knowledge of the physical world through sensations. Critics of traditional foundationalism complain that it is only through sheer repetition of the positions that many philosophers got used to talk about inferring the existence of a table from propositions about the character of sensation, or inferring propositions about the past from propositions describing present consciousness. Such critics have correctly pointed out that if these claims are intended to be phenomenologically accurate descriptions of our epistemic relation to the world they are hardly credible. Anyone who has tried to paint knows that it is very difficult to distinguish the world as it appears from the world as it is. That there is a conceptual distinction between phenomenological appearance and reality seems to me obvious. If the difficulty of artistic representation shows that we rarely reflect on appearances (as opposed to reality) it also seems to show that there is such a thing as appearance. But whether or not one can introduce an intelligible "noncomparative"15 use of "appears," it does not alter the phenomenological fact that we do seldom, if ever, consciously infer propositions about the physical world from propositions describing the character of sensation.

We also seldom consciously infer propositions about the past from anything we might call a memory "experience." The very existence of memory "experience" is far from obvious. We might also observe, however, that our commonplace expectations about the future are often not formed as a result of careful consideration of premises describing past correlations of properties or states of affairs, When I expect my next drink of water to quench my thirst instead of killing me, I do not first consider past instances of water quenching thirst, It is useful to reflect carefully on this fact for even most naturalists will view this kind of knowledge as involving inductive inference. We must, therefore, be cautious in reaching conclusions about the role of phenomenology in determining whether a justification is inferential or not. We must distinguish questions about the causal origin of a belief from questions about the justification available for a belief.


We must also distinguish between occurrent and dispositional belief. It may be that I have all sorts of dispositional beliefs that are causally sustaining my beliefs when I am completely unaware of the causal role these dispositional beliefs play. In introducing this discussion I suggested that it was ironic that naturalists would find attractive the fact that their naturalistic epistemology can accommodate the apparent phenomenological fact that far less commonsense belief involves inference than is postulated by traditional foundationalism. The irony is that phenomenology should have no particular role to play for our paradigm naturalists in reaching conclusions about what is or is not inferentially justified. According to these naturalists, the epistemic status of a belief is a function of the nomological relations that belief has to various features of the world. These nomological facts are complex and are typically not the kinds of facts which have traditionally been thought to be under the purview of phenomenology. I suppose a reliabilist, for example, can define some belief- producing process as "phenomenological." But again, even if one can describe such a process it will be a contingent question as to what beliefs such a process might justify, a contingent question that goes far beyond the competency of most philosophers qua philosophers (and certainly most phenomenologists qua phenomenologists) to answer.


But perhaps I am being unfair in suggesting that our paradigm naturalistic epistemologists have no particular credentials qualifying them to assess the question of whether the skeptic is right or wrong in denying the availability of noninferential justification for beliefs under skeptical attack. The skeptics, after all, had arguments in support of their conclusion that we have no noninferentially justified beliefs in propositions about the physical world, the past, the future, other minds, and so on. The naturalists can at least refute those arguments based on their a priori reasoning about the correct metaepistemological position. The most common way of supporting the conclusion that we do not have noninferentially justified beliefs about the physical world is to point out that we can imagine someone having the very best justification possible for believing that there is a table, say, before him, when the table is not in fact there. A person who is vividly hallucinating a table can have just as good reason to think that the table exists as you do. But we can easily suppose that there is no table present before the victim of hallucination. If direct epistemic access to the table is anything like a real relation, then it cannot be present when the table is not present. But if the victim of hallucination does not have direct access to the table, and the victim of hallucination has the same kind of justification you have for thinking that the table exists (when you take yourself to be standing before a table in broad daylight), then you do not have direct access to the table either.


The reliabilist will deny the association between noninferential justification and direct access to the table. To have a noninferentially justified belief about the table's existence is to have a belief about the table produced by an unconditionally reliable belief-independent process. The victim of hallucination has (or at least might have) a belief in the table's existence produced by an unconditionally reliable belief-independent process. It depends in part on how we define the relevant process. But if we think of the stimuli as something like sensations (which the hallucinator has), and the process as what goes on in the brain when sensation is assimilated and turned into representation. There is no reason why someone hallucinating cannot satisfy the conditions for having a noninferentially justified belief, assuming of course that the process in question really is unconditionally reliable. The reliabilist's metaepistemology allows at least a conditional response to the skeptic's attack. More precisely, the reliabilist can point out that a reliabilist metaepistemology entails that the skeptic's conclusion about the noninferential character of belief about the physical world does not follow. And, of course, everything the reliabilist says about the physical world applies to the past, other minds, and even the future. The reliabilist probably would not claim that beliefs about the future are noninferentially justified but the reliabilist should claim that there is no reason in principle why they could not be, and should continue to assert that the skeptic has no argument for the conclusion that we have no direct, i.e., noninferentially justified, beliefs about the future.


Interestingly, not all naturalists will reject the skeptic's claim about noninferential justification in the same way. Consider again the reliabilist's response to the argument from hallucination as a way of establishing that we have no noninferentially justified beliefs about the physical world. The crucial move for the reliabilist was to deny that we are forced to regard the hallucinatory situation as one in which the subject lacked a noninferentially justified belief. A causal theorist about direct knowledge, like Armstrong, might admit that in hallucinatory experience we lack noninferential knowledge, but continue to assert that in veridical experience we have such knowledge. This naturalist is more likely to deny the skeptic's presupposition that we should say the same thing about the nature of the justification available to the victim of vivid hallucination and the person who has qualitatively indistinguishable veridical experiences. On one version of internalism, the internalist holds that the conditions sufficient for justification are always states internal to the subject. If sensations are not themselves relations (a controversial claim to be sure), and the sensory evidence of S and R is indistinguishable, and there is nothing else "inside their minds" to distinguish their epistemic states, then this internalist will insist that if the one has a certain kind of justification for believing something, then so does the other. But a causal theorist thinks that the relevant question that determines the nature of the justification available for a belief involves the origin of the belief. The internal, i.e., nonrelational, states of S and R can be qualitatively indistinguishable but S's internal states can result in S's having a noninferentially justified belief by virtue of their being produced in the appropriate way. R 's internal states might produce the very same belief, but because they were not caused by the appropriate facts, they will not result in the having of a noninferentially justified belief. In short the hallucinator's belief cannot be traced via sensation back to the fact about the world that would make the belief true. The person lucky enough to have veridical experience typically has a belief that can be traced back to the fact that makes the belief true. This is a perfectly clear distinction, and there is nothing to prevent an epistemologist from arguing that this just is the distinction that determines whether or not someone has a justified or rational belief. Furthermore the question of whether the justification is inferential has only to do with the kinds of links in the causal chain leading to the relevant belief. If the causal connection goes directly from some fact about the physical world, to the occurrence of sensory states, to representations about the physical world, there are no other beliefs that crucially enter the story. The justification that results will be justification that does not logically depend on the having of other justified beliefs.16 It will be noninferential justification. So again, we can see how naturalistic metaepistemology can put one in a position to claim that the skeptic has not established the crucial premise concerning the inferential character of our belief in the propositions under skeptical attack.


Even if naturalism allows one to point out that the skeptic has not established the crucial first premise of the argument, it does not follow, of course, that the naturalist has given any positive reason to suppose that the skeptic is wrong in claiming that the propositions under skeptical attack are not the objects of noninferentially justified belief, If skeptics and nonskeptics are to play on a level playing field, there is no "burden of proof' when it comes to fundamental issues in epistemology. If the philosopher wants to claim that we have noninferentially justified belief in certain propositions, then the philosopher can give us good reasons to think that such justification exists. The skeptic who wishes to deny that we have such justification can give us good reasons to think that it does not exist. The skeptic, however, also has a fall-back position. Without arguing that we have no noninferentially justified beliefs in propositions about the physical world, the past, other minds, and the future, the skeptic can move "up" a level, and deny that we have any good reason to believe that we have noninferentially justified belief about the existence of noninferential justification for these beliefs. An "access" internalist7 can move from the proposition that we have no justification for believing that we have a noninferentially justified belief that P to the conclusion that we do not have a noninferentially justified belief that P, But the externalist rejects just such an inference, Even if we abandon strong access internalism, however, we might find skepticism that maintains that we have no justification for believing that we have a justified belief that P just as threatening as skepticism that concludes that we are unjustified in believing P. Before we consider the question of whether skepticism will arise at the next level up within naturalistic epistemology, let us briefly discuss the naturalist's approach to skeptical issues involving inferential justification.





Most of the general observations made about the naturalist's response to skeptical challenges concerning the class of noninferentially justified beliefs will apply as well to inferential justification. If the skeptic were to succeed in convincing the naturalist that we are not noninferentially justified in believing propositions about the physical world, for example, the naturalist will presumably argue that such beliefs are inferentially justified. The reliabilist, for example, would argue that if our beliefs about the external world result from input that includes beliefs about the internal and external conditions of perceiving, or even beliefs about the qualitative character of sensation, the relevant belief-dependent processes are conditionally reliable and therefore produce (inferentially) justified beliefs provided that the input beliefs are themselves justified. The proviso is crucial, of course, and reminds us that to establish that first-level skepticism is false, the reliabilist who concedes that the justification is inferential in character must establish the existence of at least one unconditionally reliable process and at least one conditionally reliable process.


We noted in discussing the naturalist's views about noninferentially justified belief that the naturalistic epistemology has a potentially significant advantage in dealing with skepticism precisely because there are no restrictions on how large the class of noninferentially justified beliefs might be. As I indicated, there is no a priori reason for the naturalist to deny even that we have noninferentially justified beliefs about the past and the future. Evolution might have taken care of us rather well when it comes to reaching true conclusions about the world, and evolution might have accomplished this end without burdening our brains with too many conditionally reliable belief-forming processes. Nozick's tracking relations can in principle hold between any fact and any belief, and the tracking relations need not involve any intermediate beliefs.


Just as the naturalist's class of noninferentially justified beliefs can be very large in comparison to those recognized by traditional foundationalists, so the class of inferences recognized as legitimate by the naturalist can be equally large. Consider again the reliabilist's position. There are no a priori restrictions on how many different kinds of conditionally reliable belief- dependent processes there might be. Valid deductive inferences are presumably the paradigm of a conditionally reliable belief-dependent process. Classical enumerative induction may satisfy the requirements as well provided that we find some suitably restricted characterization of the inductive "process" that succeeds in denoting and that takes care of grue/green riddles of induction.8 I suspect most naturalists will be reluctant to include perceptual beliefs among the beliefs produced by belief-dependent processes, but there is no reason why a reliabilist could not be a sense-datum theorist or an appearing theorist who holds that we do have at least dispositional beliefs about the qualitative character of sensation and who further holds that such beliefs are processed by conditionally reliable belief-dependent processes that churn out commonsense beliefs about the physical world. In short, take any kind of inference that people actually make and the reliabilist could hold that it is a conditionally reliable belief-dependent process. All one needs to do is to formulate a description of the process that takes the beliefs one relies on as premises (the input) and produces the beliefs that constitute the conclusion (the output). The description will have to be such that we succeed in picking out a kind of process that does play the causal role described, but it will not need to involve any reference to the "hardware" of the brain. Indeed we can try to denote the relevant process by directly referring only to the kind of premises and conclusion with which it is associated. Roughly the idea is that we can try to denote a belief-dependent process X, for example, using the description "the process (whatever it is) that takes premises like these and churns out conclusions like this." Of course, such a description is probably too vague to do the trick. The locution "like these" can hardly be said to characterize precisely enough a class of premises. One would need to characterize the relevant points of similarity to have a well- defined class of premises which could then enter into the definite descriptions denoting the process that takes them as input.


If we consider any argument someone actually makes there will be indefinitely many classes of propositions to which the premises and the conclusion belong, and that will enable us to formulate any number of different descriptions of belief-forming processes. This is not a difficulty for the reliabilist, for as long as we have a locution that succeeds in denoting a process playing a causal role, we can use counterfactuals to define conditional reliability. The fact that a single inference might be subsumed under a number of different reliable belief-dependent processes is hardly a problem. If the inference can be subsumed under the description of both a reliable and an unreliable process, the crucial question will be the process that is causally determining the production of a belief. Thus if someone trustworthy tells me today that it rained in New York, I can describe this as a case of processing testimony to reach a conclusion about the truth of what is testified to, or I can describe it as a case of taking a statement I hear involving the name "New York" and believing all of the noun clauses containing that name. The former, let us suppose, is a reliable belief- dependent process while the latter is not. But you will recall that in formulating descriptions of processes appealing to kinds of premises and conclusions we are merely hoping to denote some process (presumably a complex brain process) that does take input and causally produce output beliefs. It does not follow, of course, that every definite description we formulate will succeed in denoting. In the hypothetical situation we are discussing, it may be that there is no programming in the brain that takes the "New York" input and processes it in the way described. If there is nothing denoted by the description playing the relevant causal role then we do not need to worry about the fact that such a process if used would be unreliable. And we do not need to worry about the fact that we describe the inference in question as if it involved a belief-forming process of the sort described. If someone is programmed in such a way that he sometimes makes the legitimate "testimony" inference and sometimes makes the bizarre "New York" inference, the justificatory status of the resulting belief will be a function of the process that was causally operative in this case. If both processes are operating simultaneously the reliabilist will probably need something like Nozick's conception of one method outweighing another. The justificatory status of the belief will depend on which belief-forming mechanism would have prevailed had they conflicted.


To emphasize the point made earlier. according to paradigm naturalists there are indefinitely many candidates for legitimate inferential processes. There arc no a priori restrictions on how many conditionally reliable belief- dependent processes might be operating in normal human beings. There are no a priori restrictions on how many belief-dependent tracking relations might exist between beliefs and the facts that they track. Furthermore, just as in the case of the noninferential justification, the question of which inferential processes generate justified beliefs for the naturalist will be a purely contingent fact of a sort inaccessible to most philosophers. The existence of conditionally reliable processes, tracking relations, and the like, is something that could be discovered only as a result of empirical investigation into causal relations. Philosophers are not trained to engage in this sort of empirical investigation.





Based on the observations above, I would argue that if paradigm naturalist/externalist metaepistemologies are correct, then normative epistemology is an inappropriate subject matter for philosophy. Philosophers as they are presently trained have no special philosophical expertise enabling them to reach conclusions about which beliefs are or are not justified. Since the classic issues of skepticism fall under normative epistemology, it follows that if naturalism were correct, philosophers should simply stop addressing the questions raised by the skeptic. The complex causal conditions that determine the presence or absence of justification for a belief are the subject matter of empirical investigations that would take the philosopher out of the easy chair and into the laboratory.


The realization that a good part of the history of epistemology becomes irrelevant to contemporary philosophy if we become metaepistemological naturalists might cause a good many philosophers to reconsider the view. I have always found the skeptical challenge to be fascinating and it has always seemed to me that I can address the relevant issues from my armchair. If I had wanted to go mucking around in the brain trying to figure out the causal mechanisms that hook up various stimuli with belief I would have gone into neurophysiology.


To rely on the philosopher's interest in skepticism and penchant for armchair philosophy as a rhetorical device to recruit wavering naturalists, however, might be viewed as a new low in the art of philosophical persuasion. The mere fact that philosophers have been preoccupied with a certain sort of question does not mean that they were qualified to answer it. There are all kinds of perfectly respectable candidates for misguided philosophical investigations. Many philosophers, for example, have taken the question of whether every event has a cause to be a deep metaphysical issue in philosophy. As a good Humean, I would be the first to argue that it is a purely contingent question and if one wants to know the answer to it, one should not go asking a philosopher. I am also sympathetic to the less popular view advanced by the positivists that metaethics exhausts the appropriate domain for philosophical investigation into morality. If a consequentialist analysis of right and wrong action is correct, for example, questions about what kinds of actions, or particular actions, we ought to perform are very complicated causal questions. Even if philosophy gives us some special insight into what is intrinsically good and bad (a thesis that is itself highly dubious), the question of which action would maximize that which is intrinsically good and minimize that which is intrinsically bad is the kind of question that philosophers are not particularly competent to address. The kind of person who is good at figuring out the consequences of actions if the kind of person who has extensive "worldly" experience and good common sense. Perhaps philosophers of the past who were so preoccupied with normative issues in ethics had that kind of experience, but without denigrating my profession, I would humbly submit that today's academic is not the kind of person to whom one should turn for advice in dealing with real-world problems. The ivory towers of the philosophy professor are anathema to the kind of experience one needs to reach reasonable conclusions about what the world would look like if we behaved one way rather than another.


The question of whether normative ethics is a legitimate area of philosophy is far too controversial to settle with a few glib remarks, My only interest here is to point out that the history of philosophy need not constrain us when it comes to reaching conclusions about the appropriate subject matter of philosophy. Philosophers have worried about fundamental normative questions of ethics for well over two thousand years. That need not stop a philosopher from presenting a respectable argument for the conclusion that philosophical concern with ethics ends with the successful analysis of the subject matter of moral judgments. If one wants to do normative ethics in addition to metaethics, one will need to do the kind of empirical work that contemporary philosophers have not been trained to do. Analogously, the fact that philosophers have been preoccupied with the skeptical challenge for literally thousands of years should not stop contemporary epistemologists from entertaining the thesis that the appropriate subject matter of epistemology ends with metaepistemology. After the metaepistemological analysis is complete, the naturalist might argue, the only way to answer normative questions in epistemology is to engage in the kind of empirical investigation that contemporary philosophers have not been trained by philosophy to do.


I suggested in some of my introductory comments that contemporary metaepistemological debate has the potential to change the very face of the philosophical study of epistemology. Specifically, I want to argue that if any of the paradigm naturalistic accounts of knowledge and justification are correct, we should dismiss all skeptical inquiry as irrelevant to the subject matter of philosophy. If the naturalist is right, philosophers should stop doing normative epistemology.


In reaching this conclusion I should be careful to admit that the naturalistic epistemologist can, of course, embed normative epistemological conclusions in the consequents of conditional assertions. One can talk about what one would be justified in believing were certain conditions to obtain. But these conditionals are still part of metaepistemology. Indeed, such conditionals arc merely a way of illustrating the consequences of metaepistemological positions as they apply to particular hypothetical situations. A Nozick can, for example, discuss what one would or would not know about the external world if a tracking analysis of knowledge were correct and if our beliefs about the physical world track the facts that would make them true. Nozick's analysis of knowledge also has the interesting feature that we can apparently determine a priori that we do not know certain things, for example, that we do not know that there is no evil demon deceiving us. But there will be no positive normative claim with respect to empirical knowledge that Nozick is particularly competent to make qua philosopher. As we shall see in a moment, naturalism does not prevent a philosopher from reaching rational conclusions about what one is justified in believing. My conclusion is only that a philosopher's philosophical expertise is nothing that would help in reaching such conclusions. To illustrate this claim more clearly, let us turn to the question of whether naturalistic epistemology suggests that one should be a skeptic about whether or not one has justified belief.





It is tempting to think that naturalistic analyses of justified or rational belief and knowledge simply remove one level of the traditional problems of skepticism. When one reads Quine, Goldman, Nozick, Armstrong, or Dretske one is surely inclined to wonder why they are so sanguine about their supposition that our commonsense beliefs are, for the most part, justified, if not knowledge. When Nozick, for example, stresses that interesting feature of his account allowing us to conclude consistently that we know that we see the table even though we do not know that there is no demon deceiving us, we must surely wonder why he is so confident that the subjunctives which on his view are sufficient for knowledge are true. Perception, memory, and induction may be reliable processes in Goldman's sense, and thus given his metaepistemological position we may be justified in having the beliefs they produce, but, the skeptic can argue, we have no reason to believe that these processes are reliable and thus even if we accept reliabilism, we have no reason to conclude that the beliefs they produce are justified.


In the previous section I have emphasized that if naturalistic epistemology is correct then philosophers qua philosophers may not be particularly competent to answer normative questions in epistemology. I did not assert that if the view is true we have no reason to believe that we have justified belief in commonsense truths about the world around us. According to naturalistic epistemologies it is a purely contingent question as to what kinds of beliefs are justified. The existence of justified beliefs depends on nomological features of the world--facts about the reliability of belief-producing processes, the existence of tracking relations, causal connections between facts and beliefs, and the like. There are no apriori restrictions on what one might be justified in believing. But it follows from this that there are also no a priori restrictions on second-level knowledge or justified belief, It will also be a purely contingent question as to whether we have knowledge of knowledge or justified beliefs about justified beliefs. If we accept the naturalist's metaepistemological views it may be true that not only do we know what we think we know, but we know that we know these things. Similarly we may have not only all the justified beliefs we think we have, but we might also be justified in believing that we have these justified beliefs. The processes that yield beliefs about reliable processes may themselves be reliable. The beliefs about the truth of the subjunctives that Nozick uses to define first-level knowledge might themselves be embedded in true subjunctive conditionals that, given the metaepistemological view, are sufficient for second-level knowledge. My belief that my belief that P tracks the fact that P might track the fact that my belief that P tracks the fact that P. And there is no greater problem in principle when we move up levels. A reliable process might produce a belief that a reliable process produced the belief that my belief that P was produced by a reliable process. There might be a tracking relation tracking the tracking relation that tracks the fact that my belief that P tracks the fact that P. To be sure the sentences describing the conditions for higher levels of metajustification might look more like tongue twisters than metaepistemological analyses but, as ugly as they are, they are perfectly intelligible, and there is no a priori reason why the conditions required for higher-level justified belief and knowledge might not be satisfied.


It is also important to note that, according to most naturalists, in order to be justified in believing that I have a justified belief that P I need not know anything about the details of the nomological connections sufficient for knowledge. Consider again reliabilism. In order to be justified in believing that my belief that P is produced by a reliable process, I do not need to know the physiological details of the brain states linking stimuli and belief, I would need to believe that there is some process producing the belief and I would need to believe that the process is reliable, but I would not need to know very much about what that process is. As I indicated earlier, one can denote the processes that produce beliefs using definite descriptions that refer directly only to the kinds of premises and conclusions that are linked by the process. Of course the definite descriptions might fail to denote, and the beliefs in propositions expressed using such definite descriptions will either be false or meaningless (depending on what one does with the truth value of statements containing definite descriptions that fail to denote). But the descriptions might be successful, and in any event the belief that there is a reliable process taking stimuli S and resulting in belief P might itself be produced by a reliable process.


All this talk about what would be in principle possible given a naturalistic metaepistemology is fine, the skeptic might argue. But how exactly would one justify one's belief that, say, perception or memory are reliable processes? The rather startling and, I think disconcerting, answer is that if reliabilism, for example, is true. and if perception happens to be reliable, we could perceive various facts about our sense organs and the way in which they respond to the external world. Again, if reliabilism is true, and if memory is reliable, we could use memory, in part, to justify our belief that memory is reliable. You want a solution to the problem of induction. There is potentially no difficulty for the naturalistic epistemologist. If reliabilism is true, and if inductive inference is a conditionally reliable belief-dependent process, then we can inductively justify the reliability of inductive inference. Our inductive justification for the reliability of inductive inference might itself be reliable and if it is that will give us second-level justification that our inductive conclusions are justified. A solution to the problem of induction will be important because with induction giving us inferentially justified conclusions, we can use inductive inference with the deliverances of perception and memory to justify our belief that those processes are reliable. I can remember, for example, that I remembered putting my keys on the desk and I can remember the keys being on the desk. If memory is an unconditionally reliable belief-independent process, then both my belief that I remembered putting the keys on the desk and my belief that I put the keys on the desk will be justified. I now have a premise that can be used as part of an inductive justification for memory being reliable. The more occasions on which I can remember memory being reliable the stronger my inductive argument will be for the general reliability of memory.

The skeptic could not figure out how t~) get from sensations to the physical world. Assume that perception is itself a belief-independent unconditionally reliable process. Assume also that whatever perception involves its specification involves reference to sensation, and assume further that we have "introspective" access to sensation. Introspective access might itself be another belief-independent, unconditionally reliable process. Given these suppositions. if reliabilism is true, then introspection can give us justified belief that we are perceiving, and perception can give us justified belief that a physical object is present. The two reliable processes together can furnish a premise that when combined with others generated in a similar fashion give us inductive justification for believing that perception is reliable. So if both introspection and perception happen to be reliable there seems to be no great obstacle to obtaining justified belief that they are reliable. Second-level justified belief is not much more difficult to get than first-level justified belief.


The above is quite in keeping with Quine's original injunction to naturalize epistemology.19 Quine suggested that we give ourselves full access to the deliverances and methods of science when it comes to understanding how we have knowledge of the world around us. Contemporary naturalists have simply given us more detailed metaepistemological views which allow us to rationalize following Quine's advice. If the mere reliability of a process, for example, is sufficient to give us justified belief, then if that process is reliable we can use it to get justified belief wherever and whenever we like.


None of this, of course, will make the skeptic happy. You cannot use perception to justify the reliability of perception! You cannot use memory to justify the reliability of memory! You cannot use induction to justify the reliability of induction! Such attempts to respond to the skeptic's concerns involve blatant, indeed pathetic, circularity. Frankly, this does seem right to me and I hope it seems right to you but if it does then I would suggest that you have a powerful reason to reject naturalistic epistemology. I would suggest that, ironically, the very ease with which these naturalists can deal with the skeptical challenge at the next level betrays the ultimate implausibility of the view as an attempt to explicate concepts that are of philosophical interest. If a philosopher starts wondering about the reliability of astrological inference, the philosopher will not allow the astrologer to read in the stars the reliability of astrology. Even if astrological inferences happen to be reliable, the astrologer is missing the point of a philosophical inquiry into the justifiability of astrological inference if the inquiry is answered using the techniques of astrology. The problem is perhaps most acute if one thinks about first-person philosophical reflection about justification. If I really am interested in knowing whether astrological inference is legitimate, if I have the kind of philosophical curiosity that leads me to raise this question in the first place, I will not for a moment suppose that further use of astrology might help me find the answer to my question. Similarly, if as a philosopher I start wondering whether perceptual beliefs are accurate reflections of the way the world really is, I would not dream of using perception to resolve my doubt. Even if there is some sense in which the reliable process of perception might yield justified beliefs about the reliability of perception, the use of perception could never satisfy a philosophical curiosity about the legitimacy of perceptual beliefs. When the philosopher wants an answer to the question of whether memory gives us justified beliefs about the past, that answer cannot possibly be provided by memory.


Again, if one raises skeptical concerns understanding fundamental epistemic concepts as the naturalist does, then there should be no objection to perceptual justifications of perception, inductive justifications of induction, and reliance on memory to justify the use of memory. If we are understanding epistemic concepts as the reliabilist suggests, for example, then one can have no objection in principle to the use of a process to justify its use. After all, paradigm naturalists are access externalists and explicitly deny the necessity of having access to the probabilistic relationship between premises and conclusion in order to have an inferentially justified belief. The mere reliability of the process is sufficient to generate justified belief in the conclusion of an argument. There is no conceptual basis for the reliabilist to get cold feet when epistemological questions are raised the next level up. Either reliability alone is sufficient or it is not. If it is, then it is sufficient whether one is talking about justification for believing P or justification for believing that one has a justified belief that P.


The above objection to paradigm naturalistic epistemologies can be easily summarized. If we understand epistemic concepts as the naturalist suggests we do, then there would be no objection in principle to using perception to justify reliance on perception, memory to justify reliance on memory, and induction to justify reliance on induction. But there is no philosophically interesting concept of justification or knowledge which would allow us to use a process to justify the legitimacy of using that process. Therefore the naturalist has failed to analyze a philosophically interesting concept of justification or knowledge.


The objection is by no means decisive. Obviously, many will bite the bullet and happily embrace Quine's recommendation to naturalize epistemology. If the argument convinces anyone it will be those who were initially inclined to suppose that paradigm naturalists and externalists will inevitably encounter skepticism at the next level up. Maybe we have knowledge or justified belief as the naturalist understands these concepts, some would argue, but we would never be in a position to know that we have knowledge or justified belief if the naturalist is right. The only reason I can see for granting the first possibility, but denying the second, is that one is implicitly abandoning a naturalistic understanding of epistemic concepts as one moves to questions about knowledge or justification at the next level. But if when one gets philosophically "serious" one abandons a naturalist's understanding of epistemic concepts, then for philosophical purposes one should not even concede the naturalist's understanding of epsitemic concepts at the first level. Once you concede that according to the naturalist we might have knowledge or justified belief about the past and the external world, you have also implicitly conceded that we might have knowledge that we have such knowledge, justified belief that we have such justified belief. And we might also have knowledge that we have knowledge that we have knowledge, and have justified beliefs that we have justified beliefs that we have justified beliefs. It seems to many of us that naturalistic epistemology is simply missing the point of the philosophical inquiry when their analyses of epistemic concepts continue to be presupposed as the skeptical challenge is repeated at the metalevels. But the only explanation for this is that the naturalist's analysis of epistemic concepts never was adequate to a philosophical understanding of epistemic concepts.




1. I refer, of course, to causal theories of reference which take the items capable of referring to include both linguistic and mental entities.

2. Armstrong (1973).

3. See, for example, Armstrong (1983) and Fales (1990).

4. Quine (1969).

5. I am trying to avoid here a detailed characterization of either naturalism, externalism. or the relation between the views. I have argued elsewhere (Fumerton, 1988 and 1993) that one should distinguish a number of quite different versions of externalism, but that it is plausible to simply identify one sense of externalism with a commitment to the kind of nomological analyses of epistemic concepts that I have called here naturalistic. Other senses of externalism have more to do with "access" requirements for the conditions that constitute knowledge and justified belief.

6. By characterizing normative epistemology this way I hope that I make it clear that the label "normative" is not meant to imply that the questions can be reduced to questions of value theory. I have argued that it is a serious mistake to think that epistemology has much in common with ethics (see Fumerton 1993).

7. As I indicated above, naturalistic epistemology could be understood in terms of the vaguer suggestions that we use science to answer traditional epistemological questions. But I think it best to reserve the term for the analysis of epistemic concepts that allows one consistently to follow the suggestion.

8. An epistemically self-refuting position is one which if true would entail that one has no reason to believe that it is true.

9. In both the realist and empiricist tradition, at least some propositions about the content of one's current mental states were taken to have this unproblematic, noninferential character.

10. On Nozick's tracking analysis of knowledge (1981), beliefs can clearly track facts without relying on any intermediate beliefs and we can again call these beliefs noninferentially justified. Armstrong (1973) explicitly distinguishes basic beliefs from other sorts of justified beliefs.

11. For a defense of such a view see Fumerton (1985) and (1993).

12. For a detailed defense of this last view see Fumerton (1985).

13. I stress "qua philosopher" for there is a real danger that I will be misunderstood on this point. I will argue that the kind of naturalism discussed here is perfectly compatible with philosophers (and anyone else) having justified beliefs and having justified beliefs about whether or not they have justified beliefs. It will not, however, be their philosophical competence that yields such justification.

14. As I shall argue shortly, this claim might be misleading. In one sense the detailed character of belief-forming processes would be best discovered by neurophysiologists. But there is another sense in which anyone can form beliefs about such processes even without any detailed knowledge of how the brain works.

15. The terminology is, of course, from Chisholm (1957), 5~53.

16. It should go without saying that there may be causally necessary conditions for the existence of such noninferential justification having to do with the capacity to form other beliefs. The dependency that concerns us, however, is logical. Justification is noninferential when no other belief is a constituent of the justification.

17.By "access" internalist, I mean a philosopher who holds that X can constitute S's justification for believing P only if S has access to the fact that X obtains. Most naturalists will reject access internalism (although I have argued that it is not obvious that their naturalism prevents them from accepting the view--see Fumerton [1988]).

18. The allusion is, of course, to the problem discussed in N. Goodman (1955), chap. 3.

19. Quine (1 969), chap. 3.




Armstrong, David. 1973.        Belief Truth and Knowledge. London.

__________. 1983.              What is a Law of Nature. London.

Chisholm, Roderick. 1957.        Perceiving. Ithaca, N.Y.

Fales, Evan. 1990.               Causation and Universals. London.

Fumerton, Richard. 1985.        Metaphysical and Epistemological Problems of Perception. Lincoln, Neb.

__________. 1988.              "The lnternalism/Externalism Controversy." In Philosophical Perspectives 2: Epistemology, edited by Tomberlin. Atascadero, Calif. 1993.

__________.                 Metaepistemology and Skepticism. Unpublished manuscript.

Goldman, Alvin. 1979.        "What is Justified Belief?" In Justification and Knowledge, edited by Pappas. Dordrecht.

Goodman, Nelson. 1955.        Fact, Fiction and Forecast. Indianapolis.

Nozick, Robert. 1981.        Philosophical Explanations. Cambridge, Mass.

Quine, W. V.1969.               Ontological Relativity and Other Essays. New York.