I. EPISTEMOLOGICAL DISTINCTIONS WITHOUT APPEAL TO REASONS

It will be obvious that the first of the two commonsense distinctions I introduced above – the distinction between the status of Jack's belief that it will rain tomorrow on the basis of his soggy tea leaves, and the status of Jill's belief that it will rain tomorrow on the basis of her meteorological investigations – can easily be accommodated by the reasons eliminationist. Jill's belief is reliably formed, and, if we accept a reliabilist account of justification, it is therefore justified; Jack's belief is unreliably formed, and so it is unjustified. Those who wish to talk about reasons here are proposing an account of the substance of the distinction between beliefs such as Jack's and those such as Jill's, but there should be no question that the reasons eliminationist can recognize that there is an important distinction here, and the eliminationist has an account of it as well.

The second commonsense distinction – the distinction between Jack's belief that he is the present King of France, which he formed as a result of a bump on the head, and Jill's belief that Jack needs to see a doctor, which she formed on the basis of hearing Jack claim that he is the King of France – is importantly different. The reasons eliminationist will recognize, of course, that Jack's belief is unreliably formed, and so unjustified, while Jill's belief is reliably formed, and so justified, but this second distinction was supposed to be different from the first. After all, the defender of reasons will want to say that in our first pair of cases Jack has bad reasons while Jill has good reasons, but in the second pair of cases, Jack has no reasons at all while Jill has good reasons. If the reasons eliminationist views the second pair of cases as just another example of the way in which some beliefs are reliably formed, while others are not, then the eliminationist is unable to make a distinction which the defender of reasons, and commonsense as well, it seems, is committed to making.

The reasons eliminationist will insist, and rightly so, that it is not always an advantage of a theory that it commits us to additional distinctions. The defender of reasons sees a continuum of cases, starting with beliefs formed on the basis of excellent reasons and proceeding through weaker and weaker reasons down through cases in which beliefs are formed on the basis of pathetically weak reasons; but then, it will be insisted, there is a sharp break. There is a difference in kind, and not just in degree, according to the defender of reasons, between forming a belief on the basis of terribly weak reasons – such as Jack's belief about the weather formed on the basis of tea leaf reading – and beliefs formed for no reasons at all – such as Jack's belief that he is the present King of France. For the reasons eliminationist, on the other hand, we simply have a single continuum of cases, differing only in degree of reliability. There is no difference in kind, on this view, between Jack's belief about the weather and Jack's belief that he is royalty. Indeed, as the reasons eliminationist sees it, we should not assume that beliefs which are formed, as the defender of reasons would describe things, on the basis of no reasons at all, are automatically lower down on this continuum than any of the beliefs which are formed on the basis of weak reasons.

Notice how each of these views would describe these cases. For the defender of reasons, there is a difference in kind between forming beliefs on the basis of reasons, however weak those reasons may be, and cases like Jack's belief that he is French royalty. In the former kind of case, there is a psychological explanation for the origin of the belief in question. Jill's belief that Jack has a worrisome medical problem is caused by another belief she has: that Jack is making utterly crazy claims. Her belief about Jack's medical problem is thus a result of an inference, and Jill's psychology thus includes some inferential mechanism which takes her belief about what Jack has said as input and produces her belief that Jack has a medical problem as output. If we are reliabilists, and the defender of reasons may in fact be a reliabilist, then we may go on to assess the reliability of such an inferential mechanism. But Jack's belief that he is the present King of France is quite different; it is not formed by inference. There is, plausibly, no psychological explanation for Jack's forming this belief. Instead, we must descend to the neural level to explain how this particular belief came about, and, for the defender of reasons, this makes this case fundamentally different in kind from the former case.

The reasons eliminationist sees things differently. After all, we sometimes form beliefs, quite reliably, in ways that do not have any psychological origin. Explaining the origin of these beliefs requires that we descend to the physiological level, just as we did in the case of Jack's bump on the head. Proprioception is arguably like this. We each have beliefs about the orientation of our bodies. There is a physiological explanation of how these beliefs are produced, but that may well be all of the explanation that there is of them. Let us suppose, for the sake of argument, that such an account is correct. Beliefs produced by proprioception, we are assuming then, are not just different from inferential beliefs; they are not based on sensory experience either. What is interesting about the case of proprioception is that, unlike ordinary perceptual beliefs, there seems to be no sensory experience mediating between neural input and belief. This is, at least, what I want to assume here.

Now if this is the right account of the origin of proprioceptive beliefs, the defender of reasons will have to assimilate them to Jack's belief that he is the King of France, for these beliefs are not based on reasons of any kind; they operate merely in “the space of causes,” as Sellars would put it. And since these beliefs are not based on reasons, and the defender of reasons sees justification, and thus knowledge, as demanding a basis in reasons, these beliefs are unjustified, and so cannot constitute knowledge. But for the reasons eliminationist, the proper way to think about cases like this is quite different. We merely need to ask ourselves how reliable the process is by which such beliefs are formed. If it is a reliable process, as proprioception in normal conditions certainly is, then the resulting beliefs are justified, and typically cases of knowledge as well. From an epistemological point of view, the reasons eliminationist will insist, these beliefs are more like Jill's belief that it will rain tomorrow, and her belief that Jack has a serious problem as well, than they are like Jack's belief that he is the King of France. In order to see things properly, the reasons eliminationist will insist, we must do nothing more than look at questions of reliability, and let reasons drop out of the picture entirely. If we insist on classifying these cases as the defender of reasons would have it, we don't just add an unnecessary distinction to our epistemological picture. We do something far worse than that. By viewing these cases in terms of reasons, the reasons eliminationist will insist, we are led to misjudge the epistemological status of the beliefs under discussion.

It is no small cost to reject the possibility of justified belief and knowledge on the basis of proprioception. The defender of reasons seems forced, however, to just such a position, while reasons eliminationism treats proprioception as a source of justified belief and knowledge. This surely seems to be an advantage for the reasons eliminationist.

2. REASONS ELIMINATIONISM AND SELLARSIAN DILEMMAS

Simple perceptual beliefs raise a host of complex issues for those who put reasons at the center of their epistemological theories. If we examine the etiology of perceptual beliefs, then in the typical case, although beliefs about sensory experience do not come into play, sensory experience itself is surely found in the causal chain that leads to belief. My belief that there is a tree just outside the window in front of me is caused, in part, by a sensory experience which I currently enjoy. As Sellars famously argued, however, those who favor some sort of foundationalist account of justification face a troubling dilemma: either my experience itself should be understood as having propositional content, in which case it too, like my perceptual belief, stands in need of further justification, or the sensory experience fails to have propositional content, and can thus do no justificatory work.

Foundationalists have not been silent on this problem. They have, of course, had a great deal to say about it. Some have sought to show how propositional content may be assigned to experiential states in a principled way, and then show just how such states may play a justificatory role in perceptual belief. Others have tried to slip, somehow, between the horns of the dilemma, allowing sensory states to play a justificatory role even without assigning them any propositional content. Still others have attempted to show, instead, that sensory states, when properly understood, should be seen as playing a causal role in the production of perceptual belief, without thereby playing any justificatory role at all. But it is surely no exaggeration to say that each of these approaches faces serious problems, and it is for this reason that none of them has won general assent.

The reasons eliminationist proposes a simple solution. We should stop talking about reasons here, and, in so doing, the Sellarsian dilemma goes away. When we look at the etiology of perceptual belief, it is certainly true that sensory states seem to play an important causal role. If we wish to know about the justificatory status of these beliefs, all that is relevant is the reliability of the process which produces them. If our psychological theories tell us that there is theoretical work to be done by assigning propositional content to sensory states, then we should do that. But if, on the other hand, our best theories of the origin of perceptual belief assign no propositional content at all to sensory states, that is no problem either. Whichever way our psychological theory goes, the only issue for our theories of justification and knowledge is whether the processes which produce perceptual beliefs are reliable. And there can be little doubt that in a great many cases, such beliefs are indeed reliably produced. The relationship between sensory states and perceptual belief thus understood fails to raise the issues which give rise to the Sellarsian dilemma because we no longer seek to explain the epistemological status of these beliefs in terms of reasons. There are no puzzles about how it is that sensory states, whether they have propositional content or not, may play a crucial role in processes which reliably lead to true belief. As the reasons eliminationist sees it, the Sellarsian dilemma is the product of a mistaken picture about the nature of justification and knowledge: the view that these central epistemological notions should be explained in terms of the having of reasons.

A similar issue arises for the defender of reasons who is committed to the existence of a priori justification and knowledge. Let us suppose, for the sake of argument, that some of our beliefs are, in fact, a priori justified. Some of these beliefs, it seems, are arrived at by way of inference, and the defender of reasons will explain why it is that they have the epistemological status they do by appealing to the way in which proper inference may transmit justification. Thus, for example, if I come to believe that the Pythagorean Theorem is true by recognizing the soundness of a standard proof of it, there is an obvious way in which, it seems, we may explain the epistemological status of the resulting belief in terms of the reasons I have for believing it. But not all apparently a priori beliefs are arrived at, it seems, by way of inference. Consider a candidate for such non-inferential a priori status: the belief that nothing may be red and green all over at the same time, or the belief that the law of the excluded middle is true, or the belief that the sum of one and one is two. What, according to the defender of reasons, is the source of the justificatory status of such beliefs? A variety of accounts have been offered here, involving, among other things, appeals to intuition or intellectual seemings. Such approaches face the very same problem which the status of sensory experience gave rise to: either intuitions, or intellectual seemings, or what have you, have propositional content, in which case they stand in need of further justification themselves, or they lack propositional content, in which case they seem incapable of playing a justificatory role. Any account of a priori knowledge which explains the epistemological status of our beliefs in terms of reasons, will face just such a Sellarsian dilemma about basic a priori beliefs.

Now many epistemologists of a naturalistic turn of mind will simply reject the a priori/a posteriori distinction. It is certainly true that, under Quine's influence, the rejection of the a priori was long seen as a hallmark of a naturalistic approach to epistemological issues. But those times are now long past, and many naturalistic epistemologists have argued that there is a perfectly coherent distinction to be made among reliably formed beliefs between those which are the product of experiential input to belief-forming mechanisms, and those which are formed without any such input.Footnote ³ Naturalists who wish to preserve talk of reasons in their account of key epistemological distinctions will need to face the Sellarsian dilemma for a priori belief just as much as those who reject naturalism. But the reasons eliminationist will urge that we may avoid the dilemma entirely, just as we did in the case of perceptual belief, by doing away with talk of reasons. If some of our beliefs are reliably formed without any experiential input, then those beliefs are thereby justified and so candidates for knowledge as well. We may look to experimental work in psychology to see if there any beliefs which answer to such a description, as some philosophers and psychologists have urged is quite plausible, but the epistemological status of such beliefs raises no special problems at all. For example, if part of our native intellectual equipment includes some inferential tendencies which reliably track some basic deductive rules, and if these inferential tendencies may produce some beliefs about logic without involving any experiential input, then the reliability of the processes producing these beliefs is thereby assured, while at the same time, assuring the a priori status of those beliefs.Footnote ⁴ All of this may be easily explained without appeal to talk of reasons. As the reasons eliminationist sees it, talk of reasons here is nothing but a source of apparently unsolvable problems. It is a mistake to try to address these problems head-on, as the defenders of reasons seek to do. What we should do, instead, according to the reasons eliminationist, is reject the very picture which gives rise to these problems. We should stop trying to explain justification and knowledge in terms of the having of reasons. An epistemology without reasons allows us to do precisely that.

3. REASONS AND DOXASTIC DELIBERATION

While the Sellarsian dilemma focuses our attention on the premises from which our inferences proceed, it should not be assumed that the status of inference itself is unproblematic for the defender of reasons, if only the legitimacy of those premises can be explained. Inference itself raises problems for the defender of reasons, problems which may simply be sidestepped if we endorse reasons eliminationism.

There is a traditional picture of how it is that legitimate inference may play a justificatory role according to which our most basic inferential principles are a priori justified. Simple deductive principles, according to this picture, such as Modus Ponens, may be recognized a priori as licensing inference because they are truth-preserving. Such an account, of course, requires that we successfully address the Sellarsian Dilemma for a priori justification, but the defender of reasons had to face that problem anyway. But deductive inference, by itself, cannot possibly define the limits of legitimate inference, and on the traditional picture, there are also certain non-deductive principles, such as those involved in inductive inference, or inference to an explanation, whose legitimacy needs to be explained as well. Here too, on the traditional picture, it will be claimed that such inferences are a priori legitimate, although here, of course, this cannot be explained by appeal to their truth-preserving character, since non-deductive inference is not truth-preserving. On the standard view, it will be claimed that we may know a priori that good principles of inference are justification-conferring, despite the fact that they are not truth-preserving, and this has led to a wide variety of arguments which attempt to show, a priori, how certain non-deductive principles of inference meet this standard. The history of attempts to give such a priori defenses of various inductive principles, however, has not been salutary.

Notice that it will not do, of course, simply to show that, given one's other epistemological commitments, such principles of inference had better be a priori defensible or else an untenable skepticism will result. Although a number of philosophers have made such arguments, what is needed here is something which provides a certain amount of epistemological illumination: if we are committed to an account of justification or knowledge which explains these notions in terms of reasons, we will need to know how it is that we have good reason to make inferences which accord with the principles in question. The insistence that, if we are to reject a very broad skepticism, the legitimacy of such principles must be knowable a priori does not provide the needed illumination. To put the point slightly differently: the fact that defenders of a reasons-based approach to epistemology seem committed to showing a priori that basic inferential principles are legitimate need not be taken as an argument for such a priori legitimacy; it may, instead, be taken as an argument against the reasons-based approach itself.

Reliabilists who wish to maintain the reasons-based approach may object here. While it is certainly true that internalist epistemologists are committed to the a priori defensibility of basic principles of inference, the philosopher who wishes to reconstruct reasons talk in terms of reliability has no such commitment. A reliabilist may view principles of inference as legitimate to the extent that they are reliable, whether they are a priori defensible or not. One belief provides good reason for another just in case the second belief may be reached from the first by way of a reliable principle of inference. On such a view, good reasons may be explained in terms of reliability, and good inference as well; reliability may be seen, on such a view, as a way to resuscitate the notion of reasons, rather than a way to undermine it. Such a reliabilist will think that it is a certain traditional internalist account of reasons which needs to be given up, but we need not give up the very idea of good reasons.

Such a view certainly deserves to be taken seriously, but I believe that reasons eliminationism deserves a serious hearing as well. There are a number of considerations which suggest that a reliabilist should not very quickly commit to a reliabilist reconstruction of talk about reasons. Discussions of the epistemology of inference often begin with very simple examples, as I myself did above, involving applications of Modus Ponens. But as Gilbert Harman has long argued, Modus Ponens is not a principle of inference.Footnote ⁵ That q is entailed by p, and if p, then q, does not tell us what we should infer if we believe both p and if p, then q. It may be that we should infer q, but it may also be that we should give up either p, or if p, then q, or both. Inference, as Harman argued, is more holistic than the simple examples would suggest.

Exactly how holistic inference actually is is a difficult question about human cognition. More than this, there are difficult questions about how such belief change is to be explained. It could be that inference proceeds by way of fairly local principles of inference, where these principles are not a priori defensible, but are, instead, often reliable in the environments in which they are used. If this were the case, the effect of the workings of such inferential principles would be to produce the appearance of some holistic process – since the individual inferences would typically be reliable, and a body of truths is, of course, coherent. Something very much like coherence would thus be achieved not by way of inferential principles which are especially sensitive to coherence, but instead as a by-product of the reliability of the individual inferences. If our minds work in ways like this, then although we need to depart from the internalist picture which has our inferences licensed by their a priori defensibility, it might still be appropriate to talk of reasons for belief.

Alternatively, it might be the case that belief change is achieved by way of principles of inference which genuinely operate holistically. Exactly what such holistic inferential principles might look like is not at all clear, especially since any attempt to spell out a set of workable principles of this sort would need to be ever mindful of problems of computational complexity, but we need not suppose that any such principles would have to be a priori defensible. It would surely be good enough if we were able to discover principles of belief change of this sort which operated reliably within the kinds of environments we tend to inhabit. So long as belief change is carried out by way of inferential principles, however, whether local or more holistic, it seems quite natural to explain it in terms of reasons. Admittedly, if the actual processes of belief revision involve holistic principles of inference, then the illustrations we typically use in epistemology, involving, for example, applications of Modus Ponens, are horribly oversimplified, and the way in which we typically talk of reasons for belief is horribly oversimplified as well, but, all the same, on this picture, there seems a place for talk of reasons, rather than their wholesale elimination.

There is, however, another possibility. It may be that belief change is not achieved by way of principles of inference at all, whether a priori defensible or otherwise. Change of belief may be a product of neural processes which do not implement anything like inferential principles, but which operate, instead, by various sorts of activation systems, as defenders of parallel distributed processing have typically suggested.Footnote ⁶ While some who have defended accounts of this sort have sought to do away not only with talk of inference, but talk of belief as well, this is not universally true. On many such accounts, the notion of belief survives our understanding of cognitive change, even if the notion of inference needs to be given up. The very idea of inference, on such a view, involves a vast oversimplification of what goes on in belief change. When we see belief change for what it is, it will be argued, all talk of inference drops out of the picture, and once we give up on the notion of inference, the very idea of reasons is seen to be mistaken as well.

One need not be committed to such an account of belief change in order to think that we should adopt an epistemology without reasons. If an account of belief change without inference is even a live possibility, we should not adopt an epistemology which presupposes that it is mistaken. One might thus favor an epistemology which does not commit to the importance of the category of reasons unless and until we can be quite sure that an account of change of belief which does away with inference is mistaken. An epistemology which distinguishes good belief change from bad belief change in naturalistic terms, as reliability theories do, can remain neutral on the question of whether talk of reasons is genuinely illuminating, or, instead, simply misleading.

4. REASONS WITHOUT PSYCHOLOGICAL COMMITMENTS?

Those who wish to defend the place of reasons in epistemology may seek to avoid getting involved in discussions of cognitive architecture. The distinction between “the space of reasons” and “the space of causes,” it may seem, is extremely robust, and far more secure than any current account of the mechanisms of belief change. In order to defend the importance of this distinction, one need not hold, of course, that reasons for belief are not causes. The suggestion that there is a robust distinction here worth defending merely involves the idea that there is a distinction in kind between a reason for belief and a mere cause.

Let us return to the case introduced at the beginning of this paper. When Jack hits his head on a rock and comes to believe, as a result of his injury, that he is the present King of France, his belief is caused, but he lacks any reason for holding it. And this case seems different in kind, and not merely in degree, from cases in which Jack forms beliefs for good reasons, or even for very weak reasons. The importance of the distinction seems clearest in cases of self-conscious deliberation. Jill sees Jack lying on the ground with a slight head wound. There's some blood, and he's briefly knocked unconscious. When he comes to, he takes himself to be the King of France. And perhaps Jill wonders, “Can I safely walk down the hill with Jack and get him to a doctor, or do I need to call an ambulance immediately?” In situations like this, people like Jill may self-consciously ask themselves about the reasons they have for various beliefs. They consider the reasons on either side of an issue, and they try to figure out what to believe. There seems no other way to describe this situation without invoking the notion of a reason. More than this, such situations of doxastic deliberation seem quite central to epistemology. On some views, the most fundamental questions which epistemology can address concern the guidance we might offer to agents in these situations, and it is hard to see how we might perspicuously parse such guidance without making use of the notion of a reason for belief.Footnote ⁷

No one should dispute this account of the phenomenology of doxastic deliberation. We seem to be capable of making decisions about what to believe, and we certainly do, at times, self-consciously consider the strength of our reasons. The important questions for epistemology, however, are not about the phenomenology of doxastic decision-making. They concern, instead, questions about the nature of justified belief and knowledge; about the nature of epistemic normativity; and about the extent to which the phenomenology of doxastic decision-making actually tracks the psychology of belief acquisition and change. It is only in the light of a clear understanding of these issues that we may hope to determine whether the notion of a reason should play a central part, or, indeed, any part at all, in epistemological theorizing.

There can be no question at all, in particular, that the phenomenology of doxastic decision is deeply misleading about the actual processes that go on when we wonder about what to believe.Footnote ⁸ Our beliefs, even when we reflect, are influenced by a wide variety of factors which are unavailable to the reflective agent. Factors which seem to influence us when we reflect may, in fact, play no causal role in determining what we come to believe. Although the reflective agent seems to have some sort of direct grasp of the cognitive processes which take place in doxastic decision-making, the appearances here are deeply deceiving. Reflective agents have only a very partial view of the processes that go on during reflection, and even that partial view is not entirely accurate as far as it goes. The perspective of the reflective agent who considers what to believe is thus extraordinarily misleading. It is not even clear that we should be talking about agency here at all.

What this means, of course, is that even if the notion of reasons for belief is essential to the way in which individuals conceive of what they are doing when they are engaged in doxastic deliberation, this need tell us nothing at all about the extent to which an epistemological theory should take this notion seriously. After all, the fact that sincere participants in religious rites conceive of what they are doing in ways which make ineliminable appeal to the notion of a deity does not provide us, by itself, with a quick argument for theism. Since we are well aware that there are a variety of illusions about mental states and processes which the first-person perspective is subject to, we need to be especially careful here.

A number of factors combine to provide the doxastic deliberator with a vivid, though illusory, picture of the mental processes which lead to belief fixation. First, the capacity of working memory is quite small, and thus, when we try to bring to mind the considerations which bear in favor and against adopting a particular belief, it is inevitable that only a tiny fraction of the relevant considerations will be brought to consciousness. Second, as was briefly mentioned above, even apart from considerations of the size of working memory, many of the relevant considerations are simply inaccessible to reflection. The result, then, of self-conscious consideration of what to believe is that we end up focusing our attention on a very small, and unrepresentative, set of considerations which might bear on our doxastic decision. Although these considerations do not, in the typical case, even approximate a good deductive or inductive argument either for or against the claim which is under consideration, it is easy to turn any set of such considerations into a valid deductive argument by adding appropriate conditional beliefs. Thus, for example, if in trying to decide whether to belief p or not-p, I attempt to bring to mind the considerations which bear on this issue, and the considerations I can think of amount to q, r, and s, considerations which jointly seem to count, however, inchoately in favor of p, then I may turn these considerations into a valid deductive argument for p by accepting the claim that if q and r and s, then p. There is a good deal of psychological evidence that, in the situation described, the conditional will strike me as plausible. The act of reflection will thereby provide me with a valid deductive argument in favor of a particular conclusion. When I come to believe that very conclusion in such a situation, it thereby seems that I have formed the belief on the basis of just that argument. Talk of reasons for belief seems a natural way to characterize the influence of such considerations on the resulting belief.

The fact that the considerations which count in favor of a belief may be so easily put in the form of a valid argument, or, at a minimum, a convincing non-deductive argument, is a product of the forces which go to work when we stop to reflect on what we should believe, rather than some real insight into the considerations which move us. Our ability, under conditions of reflection, to put these considerations in the form of a tidy argument makes it seem natural, and utterly perspicuous, to regard the premises of such arguments as reasons for our beliefs, and the rules of inference which are needed to get from premises to conclusion as the ones which guide our reasoning.Footnote ⁹ But this is an utterly inaccurate view of the fixation of belief, even under conditions of reflection, and to the extent that conceptualizing this whole process in terms of reasons for belief is funded by this inaccurate view, the very idea of reasons for belief is thereby called into question.

It is tempting to think that we can settle the issue about the importance of reasons without getting into difficult questions about the mechanisms of belief fixation only because the phenomenology of doxastic decision provides us with such a vivid picture of precisely how belief fixation takes places under conditions of reflection. We know too much, however, about the processes which go on during reflection to take this phenomenology at face value. And this surely suggests that if we are to understand the place of reasons in epistemological theorizing, we cannot avoid addressing the difficult psychological questions about how belief fixation actually occurs.

5. BASING AND CAUSATION

I want to look at one last attempt to secure the place of reasons in epistemological theorizing which bypasses the kinds of psychological questions on which I have been focusing. I have been assuming throughout this paper that if my belief that q is part of my reason for believing that p – and here I am concerned with the reasons which provide my doxastic justification for believing that p, and not just the propositional justification for that belief – then my belief that q must be part of the cause, or part of what causally sustains, my belief that p. If the two beliefs are causally independent of one another, then neither can be a reason – in the doxastic justification sense of reason – for the other. It is this causal requirement that allows experimental results from psychology to get their purchase on these issues. There can be no doubt that psychological results give us real insight into the causal relations among our mental states, and well-confirmed theories in psychology are far more deserving of our credence than the deliverances of phenomenology.

But some will insist that this way of thinking of things fundamentally misunderstands the nature of reasons. As some philosophers would have it, my reflectively believing that q is my reason for believing that p is what constitutes my believing that p on the basis of q.Footnote ¹⁰ There is no further fact which my reflective belief, or as some would put it, my reflective endorsement of believing one thing on the basis of another, attempts to record. On this view, since the reflective belief is what constitutes my believing on the basis of a reason, it is not the sort of belief which even could be mistaken. To think that the psychological facts about causal relations among beliefs are even relevant to evaluating the truth of the reflective belief, on this view, simply fails to recognize what believing on the basis of reasons consists in. If this sort of view is correct, then facts about our reasons for belief are wholly insulated from any facts about the etiology of belief, and we need not worry, therefore, that the place of reason in epistemology may somehow be undermined by discoveries in empirical psychology.

Let us consider, then, a particular example. Suppose Jim is part of a faculty search committee, and he is reading over dossiers of applicants. A woman who has applied, with some undeniably strong credentials, is favored by some members of the search committee, but Jim has placed her file in the reject pile. When asked why he found her candidacy unacceptable, Jim cites a number of features of her record. These, he says, are the reasons he believes that she is an unacceptable candidate.

Now I simply want to stipulate that Jim is entirely sincere in stating his reasons. He has not passed over this file lightly, since he knew that some of his colleagues favored the candidacy of this particular woman, and he sincerely believes that the reasons he cites are the reasons for which he is dismissing her candidacy. Since he reflectively believes that these are his reasons, then on the view we are considering, this thereby constitutes his reason for believing that her candidacy is unacceptable.

Imagine that, at this point, one of Jim's colleagues brings up a number of studies in social psychology which show that when one and the same set of credentials were presented to subjects to be evaluated for faculty positions, some labeled with male names, and others labeled with female names, the male “applicants” were consistently rated more highly than the female “applicants,” despite the fact that their credentials were identical. Such studies seem to show, it has been argued, that the very items cited as reasons for the ratings given simply could not be the actual reasons for which the female “candidates” were rated lower than their male counterparts.Footnote ¹¹

When presented with studies of this sort, some people tough it out: “that may be true of the subjects in those studies,” they say, “but I'm not like that.” “I'm not moved by those kinds of things.” “I'm not prejudiced in that way.” “They haven't studied me.”

Given the robustness of the results in these studies, many of us may find such responses unconvincing. But what is interesting about the constitutive view of reasons under consideration here is that, on this view, Jim need not resort to such special pleading in order to defend the integrity of his assessments. What Jim should reply is not that he is immune to influence of the sort uncovered by the psychological research, but rather that the etiology of his belief is simply irrelevant to the issue of the reasons for which he holds his belief. The claim about Jim's reasons is simply a claim about the reasons he endorses under conditions of reflection. As such, the fact that he does endorse a certain claim about his reasons, a claim which has nothing to do with the gender of the candidate, thereby guarantees that his reasons have nothing whatever to do with the fact that the candidate is a woman.

Implausible as the various sorts of special pleading are in cases like this, the claim of the constitutivist about reasons seems even more implausible. If one wishes to define justification in terms of the having of reasons, then doxastic justification requires that the reasons for which one holds a belief meet certain standards. The fact that one sincerely believes that one's reasons are of a certain sort does not in any way assure that one's reasons are of that very sort, as cases like this clearly show. If the reasons one sincerely cites are actually good reasons for belief, but they are not the reasons which causally produced or sustained one's belief, then one is not doxastically justified. The attempt to identify reasons for belief with higher-order beliefs about one's reasons, or reflective endorsements of reasons, is extraordinarily implausible.

The attempt to insulate particular claims about one's reasons from the psychological facts about a belief's etiology cannot be sustained. Once we allow these facts to play a role in our understanding of the epistemologically relevant features of belief, however, we need to take seriously the possibility that the etiology of belief may be quite different from the way in which we pretheoretically believe it to be. As I have argued, this is not merely some logically possible situation in some world which we know to be very distant from our own. It is, instead, a way that belief fixation in our world might actually occur. Whether we should think of the epistemological status of a belief in terms of reasons at all, however, depends on these difficult issues about the origins of our beliefs.

6. CONCLUSION

Let me sum up. The view that the notion of reasons is central to epistemological theorizing is undeniably attractive. It seems, indeed, unavoidable, but that appearance of unavoidability may well be a product of the phenomenology of doxastic decision-making. This is, however, a poor basis for determining just how a defensible epistemology ought to be conceptualized. We know that phenomenology gets crucial features of psychological processes fundamentally wrong. This is no less true in doxastic decision than it is elsewhere. Any attempt to insulate central concepts of epistemology from the psychological facts seems doomed to failure. But when we take a serious look at the processes involved in belief fixation, it is no longer obvious that we will find a place for a distinction between “the space of causes” and “the space of reasons.” Giving up such a distinction would not mean that we would regard all beliefs as epistemically on a par. What it suggests, instead, is that the best way to make the epistemological distinctions which are needed may not involve the notion of a reason at all.

Even if the psychology of belief fixation does not lead us to give up the notion of reasons entirely – and I should say here that I expect that it will not – it may well lead us to modify our pretheoretical ideas about just where talk of reasons should prove to be illuminating. Exactly how large a role talk of reasons should play in epistemology, and, indeed, whether it should play any role at all, must be settled by our understanding of the psychological facts.Footnote ¹²