Introduction

In recent years the naturalistic approach to epistemology has been gaining currency. The goal of this introduction is to answer my title question and at the same time to place the essays that follow in appropriate perspective.

3.. Are the processes by which we do arrive at our beliefs the ones by which we ought to arrive at our beliefs?

Different theorists will answer these questions differently. The topic I wish to deal with here has to do with the relation among these three questions. If we to answer all of these questions, which should we deal with first? Can any of these questions be answered independently of the others, or will the answers to each constrain the range of answers we might give to those remaining? Just as different theorists disagree about answers to questions 1, 2, and 3, there are disagreements about the relations among these questions. What I want to suggest here is that what is distinctive about the naturalistic approach to epistemology is its view about the relations among these three questions.

One view, which I will label the traditional view, suggests a strategy of divide and conquer. Question 1 is to be assigned to philosophers,

question 2 to psychologists. Each of these groups is to conduct its research independently of the other. When both groups have completed their work, they must get together to answer question 3. It is permissible, of course, for philosophers and psychologists to meet prior to each group's completion of its assigned task. Such meetings will allow them to check progress on question 3. These meetings will not, however, have any effect on work questions 1 and 2. Question 1 is in the bailiwick of philosophers; question 2 in the bailiwick of psychologists; and the answer to question 3 is produced by comparing the answers to questions 1 and 2.

Most research in philosophy as well as psychology seems to be guided by the traditional view. On the philosophical side consider one kind of answer that has been offered to question 1: the coherence theory of justification. Coherence theorists hold, roughly, that in deciding whether to accept or reject any statement, one ought to consider how well it fits in with or coheres with one's other beliefs; one ought to adopt beliefs cohering with beliefs one already has. Whatever the merits of this view, it seems to have nothing to do with any possible answer to question 2. Suppose that psychologists were to discover that people actually arrive at their beliefs by some kind nonconscious mechanism that measures the coherence of candidate beliefs with the body of beliefs already held; candidates that cohere are adopted and those that do not are rejected. What bearing would this psychological theory have on the merits of the coherence theory of justification as an answer to question 1? None, it seems. How we actually arrive at our beliefs need have nothing to do with how we ought to arrive at them. By the same token if we are evaluating the merits of some psychological account of belief acquisition, a purported account of how people actually arrive at their beliefs in some situation, theories about how we ought to arrive at our beliefs seem to be irrelevant.

An analogy with ethics seems apt here. Consider the following three questions about human action:

These questions bear the same relations to each other as questions 1, 2, and 3. Moreover it seems clear that it is the job of ethical theorists

to answer question A and psychologists concerned with human motivation to answer question B. Only by comparing the results of these two independent investigations will the answer to C emerge. Note how absurd it would be for a philosopher to object to a psychological account of how people act on the grounds that action of that sort is immoral. It would be equally absurd if a psychologist were to object to a philosophical account of how we ought to act on the grounds that people do not act that way. There is a straightforward explanation of the absurdity of these challenges. The normative questions that philosophers ask are completely independent of the descriptive questions psychologists. This seems to be true not only in the case of questions A and B but also in the case of questions 1 and 2.

I do not mean to endorse the arguments of this section for the traditional view, nor do I mean to suggest that no other arguments for it are available. Instead I hope only to have presented enough of the traditional view and its motivation so that it may clearly be distinguished from a naturalistic approach to questions 1, 2, and 3.

I take the naturalistic approach to epistemology to consist in this: question 1 cannot be answered independently of question 2. Questions about how we actually arrive at our beliefs are thus relevant to questions about how we ought to arrive at our beliefs. Descriptive questions about belief acquisition have an important bearing on normative questions about belief acquisition.

There are, of course, different camps within the naturalistic approach; naturalistic epistemologists differ on how direct a bearing psychology has on epistemology. The most radical view here is due to Quine:

Epistemology still goes on, though in a new setting and a clarified status. Epistemology, or something like it, simply falls into place as a chapter of psychology and hence of natural science. It studies a natural phenomenon, viz., a physical human subject. This human subject is accorded a certain experimentally controlled input--certain patterns of irradiation in assorted frequencies, for instance--and in the fullness of time the subject delivers as output a description of the three dimensional external world and its history. The relation between the meager input and the torrential output is a relation that we are prompted to study for somewhat the same reasons that always prompted epistemology; namely, in order to see how evidence relates to theory, and in what ways one's theory of nature transcends any available evidence.¹

I will speak of the view that epistemological questions may be replaced by psychological questions as the replacement thesis.

Quine's argument for the replacement thesis in chapter 1 is this: the history of epistemology is largely the history of the foundationalist program. Foundationalists tried to show that there is a class of beliefs--typically beliefs about our own sense experience--about which it is impossible to be wrong. These beliefs were held to be sufficient to justify the rest of our beliefs; thus, in addition to identifying those beliefs that would serve as the foundation of knowledge, foundationalists sought to show how foundational beliefs provide us with good reason for adopting the remainder of our beliefs. The history of epistemology shows that the foundationalist program has faced one failure after another. The lesson to be learned from these failures, according to Quine, is not just that foundationalists had mistakenly answered question 1 in claiming that the appropriate way to arrive at one's beliefs is to begin with beliefs about which one cannot be wrong and build upon that foundation. Rather, according to Quine, foundationalists were asking the wrong questions. Once we see the sterility of the foundationalist program, we see that the only genuine questions there are to ask about the relation between theory and evidence and about the acquisition of belief are psychological questions. In this view question 2 is relevant to question 1 because it holds all the content that is left in question 1. The relation between these two questions is much like the relation atheists believe to hold between questions about God's act of creation and questions about the details of, for example, the big bang; the latter questions exhaust all the content there is in the former questions.

One illustration of the way in which traditional epistemological questions have become transformed through our newly gained understanding forms the heart chapter 3. Philosophers have long asked, how is knowledge possible? This question has stood for centuries as a request for a response to the skeptic, and the result has been the various attempts to work out the details of the foundationalist program. What we are now in a position to understand is precisely how foundationalists misinterpreted this important question. If the question about the possibility of knowledge is interpreted as a request to respond to the skeptic on his or her own terms, then any attempt to answer the question is doomed to failure. Quine argues, however, that this interpretation misallocates the very worries that gave rise to the question in the first place. It is through the rise

of science that we were first led to question the limits and possibility of knowledge. As science made clear the falsity of our former beliefs and our susceptibility to illusion, the question naturally arose as to whether the beliefs we arrive at, even under the best conditions, are likely to be true. In short the question arose as to whether knowledge is possible. Insofar as this question arises from within science, we may call on the resources of science to answer it. Far from making epistemology a necessary prerequisite to doing science, this makes epistemology continuous with the scientific enterprise.

Fortunately, when we turn to science to answer our question, we are not disappointed. As Quine suggests, "There is some encouragement in Darwin." Creatures whose belief generating mechanisms do not afford them cognitive contact with the world "have a pathetic but praiseworthy tendency to die before reproducing their kind." Since believing truths has survival value, the survival of the fittest guarantees that our innate intellectual endowment gives us a predisposition for believing truths. Knowledge is thus not only possible but a necessary by-product of natural selection.

This Darwinian argument may be thought to provide a motivation for the replacement thesis quite different from that offered by Quine in chapter 1. If the Darwinian argument can be worked out in detail, it may provide a way of tackling the original three questions by taking on question 3 first. If nature has so constructed us that our belief generating processes are inevitably biased in favor of true beliefs, then it must be that the processes by which we arrive at beliefs just are those by which we ought to arrive at them. Question 3 is thus answered with an emphatic affirmative, and we may move on to the remaining two questions. If we know in advance, however, that we arrive at beliefs in just the way we ought, one way to approach question 1 is just by doing psychology. In discovering the processes by which we actually arrive at beliefs, we are thereby discovering the processes by which we ought to arrive at beliefs. The epistemological enterprise may be replaced by empirical psychology.

Notice that the attempt to defend the replacement thesis by way of the Darwinian argument requires that the conclusion of that argument be given a very strong reading. Someone who concludes on the basis of natural selection that the processes by which we acquire beliefs must be roughly like the processes by which we ought will not be in a position to defend the replacement thesis. If psychological investigation is to be able to replace epistemological theorizing, there

must be a perfect match between the processes by which we do and those by which we ought to acquire beliefs. Without such a perfect match, the results of psychological theorizing will only give an approximate answer to question 1, and epistemology will be called on to make up the slack. Psychology would thus be strongly relevant to epistemology, and this version of the Darwinian argument would thus motivate a version of naturalistic epistemology, but the resulting view would be far weaker than the replacement thesis.²

Still a third argument for the replacement thesis is to be found in such writers as Davidson,³ Dennett,⁴ Harman,⁵ and, once again, Quine.⁶ In Harman's version, the argument is as follows:

We normally assume that there are basic principles of rationality that apply to all normal human beings. . . . We come to understand someone else by coming to appreciate that person's reasons for his or her beliefs and actions, or by seeing how that person made a mistake. Someone who reasoned in a fundamentally different way from the way in which we reason would really and truly be unintelligible to us.... In assuming, as we normally do, that we can make sense of other people, given sufficient information about them, we presuppose that everyone else operates in accordance with the same basic principles as we do.⁷

As Harman makes clear, he does not mean to be arguing only for the conclusion that we all arrive at our beliefs in the same way but rather that rational belief acquisition consists of arriving at beliefs in the way we all do. In Harman's view since individuals who reason in a way different than we do would be unintelligible to us, we would not count them as rational; the only rational individuals are thus ones who reason as we do. Once again this allows us to approach the original three questions by answering question 3 first, and in answering it in the affirmative, the way is paved for the replacement of epistemology by psychology. It should also be noted that the conclusion of Harman's argument must be interpreted in quite a strong way if it is to serve as an argument for the replacement thesis. When Harman argues that someone who reasons in a fundamentally different way would be unintelligible to us, he cannot simply mean that large differences in the way individuals reason would result in mutual uninterpretability, for this view is compatible with the claim that individuals differ from one another in minor respects in the way they reason. Psychology could then describe the different ways in which individuals reason but it would be impotent to pick out which of these ways (if any) were rational. Harman must therefore be arguing for the con-

clusion that any difference whatsoever in the ways individuals reason would result in mutual uninterpretability. Weaker conclusions than this will not support the replacement thesis.

We are now in a position to distinguish between a strong and a weak version of the replacement thesis. The strong version of the replacement thesis is argued for by Quine in chapter 1. Quine argues not only that epistemological questions may be replaced by psychological questions but also that this replacement must take place; psychological questions hold all the content there is in epistemological questions. On this view psychology replaces epistemology in much the same way that chemistry has replaced alchemy. The other two arguments examined for the replacement thesis, however--the Darwinian argument and the argument from mutual interpretability--suggest a weak version of the replacement thesis. In this view psychology and epistemology provide two different avenues for arriving at the same place. Psychology may replace epistemology because the processes psychologists identify as the ones by which we do arrive at our our beliefs will inevitably turn out to be the very processes epistemologists would identify as the ones by which we ought to arrive at our beliefs. Thus even if all epistemologists were to give up their trade and turn to auto mechanics, the questions they tried to answer would nevertheless be addressed, in a different guise, by psychologists. This view is carefully scrutinized in chapter 15 by Stephen Stich, "Could Man Be an Irrational Animal?"

There is a world of difference between the strong and weak version of the replacement thesis. The question at issue is the autonomy of epistemology. Are there legitimate epistemological questions that are distinct in content from the questions of descriptive psychology? Advocates of the strong version of the replacement thesis answer this question in the negative, advocates of the weak version answer it in the affirmative. The consequences of the strong version of the replacement thesis for the study of epistemology are clear: epistemology must go the way of alchemy and be absorbed into another science. In this section I examine some of the consequences of the weak replacement thesis for epistemological theorizing. If epistemology is an autonomous discipline but subject nevertheless to the constraints of the weak replacement thesis, what will epistemology look

like? What kind of relationship does this dictate between psychology and epistemology?

If the weak replacement thesis is true, epistemologists need not fear that they will be replaced by descriptive psychologists. The weak replacement thesis is a two-way street. If psychologists and epistemologists inevitably will receive the same answers to the different questions they ask, psychologists are just as subject to replacement by epistemologists as epistemologists are by psychologists. We may leave the unemployment issue aside, however, for if the weak replacement thesis is true, no kind of replacement is likely to go on. Either discipline could replace the other, but there are extremely good pragmatic reasons why they should not.

If the thesis under discussion is true, the psychology of belief acquisition and epistemology are two different fields, which ask different but equally legitimate questions and have different methodologies. In spite of these differences a complete (and true) psychology of belief acquisition will describe the same processes that a complete (and true) epistemology will prescribe. That these two fields will, when complete, single out the same processes of belief acquisition does not, however, suggest that at stages short of completion, the processes singled out by philosophers and those singled out by psychologists will match perfectly. Indeed it is clear that if pursued independently of one another, anything short of a complete psychology would look very different from anything short of a complete epistemology, in spite of their ultimate convergence. Because the two fields deal with different questions and because these questions are approached with different methodologies, processes that are easily identified by psychologists as ones that occur in us may not easily be identified by philosophers as ones that ought to be used; by the same token processes easily identified by philosophers as ones we ought to use need not be easily identified by psychologists as processes we actually make use of. The upshot is that even if the weak replacement thesis is true, no actual replacement can occur until each field has completed its work. Moreover, in order to hasten progress, philosophers and psychologists ought to be eagerly examining each other's work. If philosophers correctly identify some process as one by which we ought to arrive at our beliefs, psychologists will thereby know, even if they have not independently discovered it, that it occurs in us. Similarly if psychologists identify some process as one that occurs in us, epistemologists can be confident that this is a process by which we

ought to arrive at our beliefs, even if they have not yet reached that conclusion independently. Thus, if the weak replacement thesis is true, we can look forward to rapid progress in both psychology and epistemology as a result of their interaction, rather than either field being co-opted by the other.

Psychologism is the view that the processes by which we ought to arrive at our beliefs are the processes by which we do arrive at our beliefs; in short it is the view that the answer to question 3 is "yes." If the weak replacement thesis is true, then so is psychologism. Nevertheless it may be that psychologism is true and yet the weak replacement thesis is false. In this section I explain how that might be so.

Consider Alvin Goldman's answer to question 1 in chapter 5. Goldman suggests that we ought to arrive at our beliefs by processes that are reliable, that is, by whatever processes tend to produce true beliefs. Let us assume Goldman is correct, and let us also assume, as Goldman does not, that psychologism is true. It thus follows that the processes by which we actually arrive at our beliefs are reliable. Even though, as we are assuming, a complete psychology of belief acquisition would describe the same processes Goldman's theory prescribes, a completed psychology would look nothing like a completed Goldman-style epistemology. It would not do for a psychologist to say merely that the processes by which we arrive at our beliefs are reliable and leave it at that. Obviously a complete psychological theory must be far more detailed and specific. The added specificity of the psychological account is not merely unnecessary, however, to answering the epistemological question, if anything like Goldman's theory is correct; the psychological theory is simply spelled out in terms at the wrong level of generality to do epistemological work. Imagine that we have a complete list of all the belief acquisition processes that take place in human beings. This list, it seems, will not answer our epistemological question (even assuming psychologism to be true), for we wish to know what all these processes have in common in virtue of which we ought to acquire beliefs by way of them. Our psychological theory will not answer this question. By the same token our philosophical theory, even if given in full detail, will not answer all our psychological questions.

I have used Goldman's theory only by way of illustration; what I say about Goldman's account is doubtless true of many rival accounts as well. If a proper epistemological theory must be cashed out at a different level of generality than psychological theories, then even if psychologism is true, the weak replacement thesis is false; no replacement could ever occur, even in completed theories, if they are not couched in terms of the same generality. Psychologism is thus still weaker that the weak replacement thesis.

What implications does psychologism have for the relation between philosophy and psychology? If psychologism is true and the weak replacement thesis false, we will not be able to read our epistemology directly from our psychology nor psychology directly from epistemology. In spite of this there will be significant constraints cast on each theory by the other. If our epistemological theory tells us that we ought adopt only beliefs arrived at by processes that have a certain property, then we know that our psychological theory must attribute to believers only belief acquisition processes with that property. If our psychological theory isolates a number of different processes of belief acquisition, then an epistemologist would do well to consider what these processes have in common. This mutual readjustment will allow each discipline to advance at a more rapid rate than it would were it to proceed independently of the other. We may thus look forward to a long and fruitful relationship between philosophy and psychology.

Many philosophers who reject psychologism nevertheless believe that the processes by which we arrive at beliefs are at least roughly like the processes by which we ought to arrive at our beliefs; the one set of processes is in the same ballpark as the other. I will speak of this view as ballpark psychologism. It will be difficult to give a precise statement of this view; it is, for example, compatible with the suggestion that some of the processes by which we arrive at our beliefs are nothing like processes by which we ought to arrive at our beliefs. It is also compatible with the view that different people arrive at their beliefs in different ways. In spite of this vagueness it will be clear that this view has important implications for the relation between philosophy and psychology.

Anyone who rejects skepticism should embrace ballpark psychologism. To know something, it seems, is to arrive at a true belief in the way one ought, or, at the very least, in a way very much like the way one ought. If most people know a great many things, then many of their beliefs are arrived at by processes at least roughly like the processes by which they ought to arrive at their beliefs. In short antiskepticism implies ballpark psychologism.

If ballpark psychologism is true, there may be a fruitful interaction between epistemology and psychology, although the connection between the two will be weaker than that just outlined for psychologism. To the extent that psychologists are successful in describing some of the processes by which paradigm cases of knowledge are produced and to the extent that epistemologists are successful in describing certain general features of some belief acquisition processes, useful interaction between the disciplines will be likely.

Some are bound to object to this conclusion. I anticipate the following sort of objection:

You insist that once we reject skepticism we are committed to the mutual relevance of psychology and epistemology. Suppose epistemological questions can be answered a priori; that is, suppose it is possible to figure out, independent of any experience, which the processes are by which we should arrive at our beliefs. In this case, there would be no reason for epistemologists to consult with psychologists, for psychological work is unnecessary to reaching epistemological conclusions. There may be some sense in which psychology is relevant to epistemology, but there is no sense in which epistemologists must consult with psychologists in order to answer the questions the wish to have answered.

I want to deal with this objection. Most naturalistic epistemologists will reject the suggestion that anything is knowable a priori.⁸ I do not wish to enter that debate here. Rather I will argue that the issue of a priority is a red herring. Whether the answers to epistemological questions can be known a priori in principle, epistemologists would do well to consult psychologists in practice.

It will be useful to begin by considering an example from the theory of probability.⁹ Suppose I decide to start an insurance company and after lengthy actuarial calculations determine the rates to charge for various policies. I determine the rates in the following way. First, I write out the part of the probability calculus relevant to my problem. Then I gather information about mortality rates. By combining the abstract statements of the probability calculus with the data about

mortality rates, I can determine how much I need to charge for policies to make a profit. With this information in hand, I go about the business of selling policies.

After many years I find that my company is losing large sums of money. It may be that I am simply unlucky. It may also be that I have made mistakes in my attempt to gather information about mortality rates. There is, however, a third possibility: I may have mistakenly formulated the theory of probability. It would be foolish for me to ignore this third possibility.

Now the theory of probability is a priori knowable if anything is.¹⁰ I may be led, however, to revise my formulation of the probability theory by empirical tests. Although it may be true that in the absence of this or any other experience, a priori reasoning alone could have straightened out my errors, the fact remains that a posteriori testing may contribute to locating my error

What is important here is not whether the theory of probability is a priori knowable but how obvious it is. A priority and obviousness do not go hand in hand. Once we recognize that in the light of the difficulty of determining some of the statements of the probability theory we are liable to err in arriving at them, we may subject our theory to test by conjoining it with obviously empirical bits of information to yield a prediction. If the prediction is falsified, it may well have been that our attempt to formulate the a priori theory was mistaken.

What is true of probability theory is also true in the theory of knowledge. Even granting for the sake of argument that in principle it is possible to answer epistemological questions a priori, epistemological truths are anything but obvious. It would thus be foolhardy not to subject epistemological theories to empirical tests. If skepticism is to be rejected, then epistemology and psychology impose significant constraints on each other. The best way to develop epistemological theories is thus to employ these constraints in a way that allows us to prod the theory along by confronting it with empirical tests.

An epistemological example is in order here. Consider again Goldman's suggestion that justified beliefs are beliefs that are reliably produced. It does not matter whether one regards this suggestion as arrived at by a priori means; in any case we may subject it to empirical test. Let us consider a paradigm case of justified belief: beliefs obtained by induction. It surely seems that everyone has a great deal of inductive knowledge; If Goldman's account is correct and we are not

skeptics about induction, we should expect that when psychologists investigate human induction, they will find that inductive beliefs are reliably produced. Chapter 12 by Richard Nisbett and Lee Ross discusses some of the processes by which inductive beliefs are formed. The processes described do not seem to be reliable. In the light of this result, we have three choices: (1) to give up our antiskepticism and simply say that beliefs formed in the way Nisbett and Ross describe are not justified, (2) to deny that Nisbett and Ross accurately describe the way in which inductive beliefs are acquired, or (3) to revise or reject reliabilism.

It would be unreasonable to refuse even to consider option 3 on the grounds that epistemological truths are a priori knowable. In the end we may have good reason for maintaining reliabilism but we must seriously entertain the possibility that we have erred. Even if we want to insist that revisions in our epistemological theories are always the product of a priori recognition of previous errors even when prompted by empirical results, it would be foolhardy to ignore this additional check on our a priori reasoning.

Goldman's theory is used only as an example. The kinds of test I have described may apply to any other epistemological theory. In particular the Nisbett and Ross results provide an interesting test for any epistemological theory.

Exactly how epistemology will look if we allow it to be shaped by psychology remains to be seen. Quine's approach to epistemology is critically examined by Jaegwon Kim in chapter 2. Ellen Markman provides some interesting psychological work on our understanding of natural kinds in chapter 4; her account here is quite different from the Quinean account provided in chapter 3. The notion of observation is placed in a naturalistic perspective by Jerry Fodor in chapter 9. Fred Dretske suggests in chapter 10 that we will receive important illumination if we draw on work in the mathematical theory of communication. Christopher Cherniak shows how work in the theory of computational complexity bears on the very form of an epistemological theory in chapter 11. Chapter 6, which I wrote, attempts to resolve the dispute between foundationalists and coherence theorists. More general arguments for the relevance of empirical work for epistemology are presented in chapter 8 by Michael Friedman. Chapter 14 by Gilbert Harman is part of a larger project: a defense of psychologism. Alvin Goldman examines the relationship between our ordinary epistemological concepts and those of a scientific epistemology in chapter

13. The prospects for improving our inductive inferences are examined by John Holland, Keith Holyoak, Richard Nisbett, and Paul Thagard in chapter 16. And Stephen Stich suggests one way in which normative concepts may be integrated into a naturalistic epistemology in chapter 17.

If the arguments of the last section are sound, any epistemologist who rejects skepticism ought to be influenced in his or her philosophical work by descriptive work in psychology. The chapters in this book are a product of this influence. My hope is that this book will encourage a careful evaluation of the prospects and problems of naturalizing epistemology.¹¹

9. For a historical discussion of probability theory in which the interaction between putatively a priori reasoning and a posteriori reasoning is beautifully illustrated, see Hacking 1975 [588].

10. I do not mean to suggest that the theory of probability is a priori knowable.

11. I have received helpful suggestions from John Heil, Philip Kitcher, Linda Polson, Fred Schmitt, George Sher, and especially Louis Loeb.