Kent Bach


What bothers people about reliabilism? It has yet to be formulated adequately, but most philosophical theories have that problem. People seem to be bothered by the very idea of reliabilism, with its apparent disregard of believers' rationality and responsibility. Yet supporters of reliabilism can't seem to understand its opponents' complaints. I believe that the conflict can be clarified, if not resolved, by drawing certain distinctions which suggest that the two sides are not really talking about the same thing. After drawing these distinctions, I will offer some positive suggestions about the relation of reasoning to reliability, suggestions which in turn will depend on certain ideas about the nature of reasoning itself.

Two Kinds of Reliabilism

First we must be clear on the kind of reliabilism at issue. Originally, as with David Armstrong's and Alvin Goldman's versions, reliabilism was a theory of knowledge. Epistemological reliabilism was designed to solve the Gettier problem not by augmenting but by replacing the justification condition in the traditional analysis of knowledge as justified true belief. The problem was to find a suitable conception of reliability. Armstrong's conception, for example, was too strong, seeming to entail that a reliably formed belief must be true. At any rate, it wasn't long before reliabilism took the form of a theory of justified belief itself, thanks mainly to Marshall Swain and again to Goldman. The idea, roughly, is that to be justified a belief must be formed as the result of reliable processes, where reliability does not entail truth. I take it that what is under dispute today is justificational reliabilism, and hereafter that will be what I mean by “reliabilism.”

The difference between epistemological and justificational reliabilism has been obscured, I believe, by two factors. The first arises when reliability is understood in terms of the notion of relevant alternatives: a process is reliable only if its use leads to the ruling out of relevant alternatives. The trouble, as no one to my knowledge has pointed out, is that what counts as a relevant alternative depends on whether we are talking about knowledge or merely about justified belief. A justificationally relevant alternative is an alternative that must be ruled out if a belief is to be justified, while an epistemologically relevant alternative must be ruled out if a justified (and true) belief is to qualify as knowledge. The believer has reason to rule out a JRA but has no reason to rule out an ERA. In any given situation, therefore, no alternative can be both an ERA and a JRA. Thus the two must be distinguished if either sort of reliabilism is to be explicated by relevant alternatives.

Confusion stems also from the fact that people on both sides of the dispute assume that the reliabilist conception is not a conception of justified belief as traditionally construed. The traditional notion is evaluative, whereas reliability seems not to be. Thus, for example, Mark Pastin refers disparagingly to the concept that reliability is supposed to help explicate as a “nouveau-justification or justification-surrogate concept.” Interestingly, Goldman seems to agree, saying that he is not retracting his previous view (epistemological reliabilism) that justification is not necessary for knowledge. He insists that he then meant classical, “Cartesian” accounts of justification. In now maintaining that justified belief is necessary for knowledge, Goldman means justified belief as understood in reliabilist terms.

Two Conceptions of Justified Belief

Nevertheless, there really do seem to be two fundamentally different conceptions of justified belief involved in the debate. By this I mean not that there are two fundamentally different accounts of what justified belief is but two different conceptions of what is to be accounted for, the internalist and the externalist conceptions. According to the first, a believer must have “cognitive grasp,” as Laurence Bonjour puts it, of whatever makes his belief justified. Being justified depends on how rational and “epistemically responsible” he is in coming to hold the belief. Whereas on the externalist conception, which reliabilism falls under, the source of justification can be “external to the person's subjective conception of the situation.” On this conception epistemic rationality/responsibility, however best explicated, is not only not sufficient but not even necessary for justified belief. It is not sufficient because a believer can reasonably and responsibly rely on false principles, and it is not necessary because, as in the case of ordinary perceptual or introspective judgments, the question of rationality or responsibility seems not even to arise.

Put most simply, externalism is simply the denial of internalism. But it can be put more informatively, if tendentiously, as holding that being justified is that property, whatever it may be, which a true, “ungettiered” belief must possess in order to qualify as knowledge. Thus put it is not committed to denying that all justified beliefs are internalistically justified. It leaves that question open. At any rate, the externalist conception of justified belief seems descriptive, whereas the internalist conception seems evaluative.

Before considering whether things are as they seem, let me point out that the issue here is not the proper standards of justification. Internalism may require that whatever makes a belief justified must be available to the believer, but it does not insist that a justified belief be certain, hence true. Like Descartes, contemporary internalists such as Bonjour affirm the first condition on justified belief, but, unlike Decartes, they reject the second. Like Goldman and unskeptical epistemologists generally, they accept a fallibilist notion of justification: a belief does not have to be true to be justified. Notice that the infallibilist conception of justified belief rules out the possibility of Gettier examples, which are supposed to show that knowledge is not merely justified true belief. A justified true belief that qualifies as knowledge in one situation might not so qualify in another, though justified in the same way and therefore, presumably, no less justified. By Cartesian standards a true belief can qualify as knowledge only if it is completely or conclusively justified, that is, so justified that it is certain and therefore cannot be false. Any candidate Gettier example would be a case of an incompletely justified belief, since there would be some relevant feature of the situation that the justification of the (true) belief failed to take into account. In such a situation the belief just happens to be true and, with its less than conclusive justification, could just as well be false.

Arguments against reliabilism

Intuitively what seems right about internalism is the idea that the epistemic merit of a belief depends on the performance of the person who arrived at it. Just as we do not downgrade an action because it has unforeseeable bad consequences, so we do not downgrade a belief because it is based on something that the believer had no reason to question. We assess his rationality and responsibility in forming the belief. Innocent ignorance does not change the assessment. This can be seen by comparing two subjectively identical situations in which a person forms the same belief on the basis of the same supposition, and yet the supposition is true in one situation but false in the other. The internalist would hold that the two beliefs are equally justified.

How can the externalist answer this internalist argument? Suppose that the pair of situations just described make up a Gettier example, so that in both situations the belief in question is true and justified. Then in the first situation the belief would qualify as knowledge, since it is based on a true supposition, whereas in the second the belief would not so qualify, since the supposition is false. Now recall that for the externalist being justified is whatever property an ungettiered true belief must have to qualify as knowledge. Other than the difference in truth value of the two underlying suppositions, there is no difference in the justifications for the two beliefs. So it seems that the externalist must allow that the belief is just as justified in the second situation as it is in the first.

However, the externalist must allow this only if the supposition in question is specific in content. For only then could it, if false, “gettier” a justified true belief. If the false supposition were a broad generalization (or a general principle), reasoning which relies on it not only cannot give one knowledge but cannot, no matter how rational and responsible the agent, yield justified beliefs. Any true, rationally/responsibly acquired belief that is based on some false generalization fails to qualify as knowledge but not because it has been gettiered. It is not the situation that keeps the belief from qualifying. The culprit is the falsity of the supporting generalization, not its local inapplicability. Thus the real conflict between internalism and externalism is that for externalism a belief's being justified is partly an empirical matter. Obviously the internalist cannot allow that relying on a false generalization could keep a rational and responsible agent from having justified beliefs. Later I will try to resolve this conflict by suggesting that although relying on a false generalization cannot give one justified beliefs, one can be justified in so doing.

In perhaps the most thoroughgoing critique of externalism, of reliabilism in particular, Laurence Bonjour charges it with violating the requirement that “beliefs that are to constitute knowledge must be epistemically ... justified ... meaning roughly that the acceptance of the belief must be epistemically rational, that it must not be epistemically irresponsible” (p. 53). He holds that any belief meeting this requirement must be based on “a justificatory argument,” hence be “inferentially justified.” Thus the requirement seems to rule out “basic beliefs,” since those beliefs, if justified at all, are not justified on the basis of other beliefs. So can they be justified at all? Consider how Bonjour diagnoses the strategy behind externalism, which he takes to be a form of foundationalism as opposed to coherentism. The strategy is to avoid the “epistemic regress of justification” by “locating a class of empirical beliefs whose justification does not depend on that of other empirical beliefs” (p. 53). The problem is that noninferentially justified beliefs obviously cannot meet the justification requirement mentioned above. There must be something that makes them justified, but it cannot be necessary for the believer himself to recognize this feature and believe it to make the belief justified. For that would render the “belief not basic after all, since its justification depends on that of these other beliefs” (p. 55). So if externalism is to solve the regress problem, “though there must in a sense be a reason why a basic belief is likely to be true, the person for whom such a belief is basic need not have any cognitive grasp of this reason” (p. 55). Bonjour goes on to give an example, which I will consider later, designed to bring out “the fundamental intuition about epistemic rationality that externalism seems to violate,” hoping if not to refute externalism then at least “to shift the burden of proof decisively to the externalist” (p. 56).

However, it is not clear that he has shifted the burden. For toward the end of his paper Bonjour admits:

Any non-externalist account of empirical knowledge that has any plausibility will impose standards for justification which very many beliefs that seem commonsensically to be cases of knowledge fail to meet in any full and explicit fashion. And thus on such a view, such beliefs will not strictly speaking be instances of adequate justification and of knowledge. But it does not follow that externalism must be correct. This would follow only with the addition of the premise that the judgments of common sense in this area are sacrosanct, that any departure from them is enough to demonstrate that a theory of knowledge is inadequate. (p. 66)

However, Bonjour is well aware that according to foundationalism (of which reliabilism is supposedly an instance), basic beliefs “provide the foundation upon which the edifice of empirical knowledge rests” (p. 54). Therefore, if he is going to deny that they really are justified and really do provide such a foundation, he needs to defend some form of coherentism, something he does not do. Until he does that, he has not shifted the burden of proof to the externalist.

Moreover, Bonjour should distinguish epistemologically basic from psychologically basic beliefs, a distinction to be supported by the conception of reasoning I will sketch later. A belief is epistemologically basic if it is justified independently of the believer's other beliefs; it is psychologically basic if not actually inferred from other beliefs. Clearly a belief can be psychologically basic without being epistemologically basic, though the converse seems false. Perhaps there are no epistemologically basic beliefs (at least no empirical ones) but plenty of psychologically basic beliefs, such as ordinary memory and perceptual beliefs. This would be so if any belief formed without inference from other beliefs could conceivably be disconfirmed by others that were brought to bear against it. In particular, as our conception of reasoning will suggest, psychologically basic beliefs can result from processes that occur only if not blocked by other processes that reliably lead to the occurrence of thoughts of reasons against the belief.

The point of mentioning the distinction between epistemologically and psychologically basic beliefs is that the reliabilist does not have to regard beliefs that are merely psychologically basic as “sacrosanct.” It is enough that they generally be justified, i.e. have what it takes generally to qualify as knowledge. Even if they are not formed by inference from other beliefs, their justification could still depend on other beliefs. So I think it is a mistake for Bonjour to assume that externalism must be motivated by the foundationalist need to escape the epistemic regress. Indeed, if justifiedness is ultimately to be explained in coherentist terms, reliabilism is a solution to a different epistemic regress problem, a problem that internalism cannot solve. This would be the problem of justifying everything on which our supposedly justified beliefs depend. Internalism cannot solve it because it treats justifiedness as a purely internal matter: if p is justified for S, then S must be aware (or at least be immediately capable of being aware) of what makes it justified and why. Reliabilism requires no such thing. Instead, it requires only that the generalizations and principles that cognitive processes follow be true in order for the beliefs that result from them are to be justified and, if true and ungettiered, to qualify as knowledge.

Justified Beliefs and Justified Believers

Curiously enough, some internalists and some externalists allow that theirs is not the only legitimate conception of justification. For example, Hilary Kornblith, an internalist, acknowledges that his notion of justified belief as the product of epistemically responsible action is not the only legitimate notion of justified belief. And Goldman the externalist distinguishes between “theoretical and regulative justification principles.” Regulative principles are for epistemic agents to follow; theoretical principles are for epistemologists to discover. Of course, this distinction would be trivial if it turned out that for every valid regulative principle there is a corresponding theoretical principle, namely one asserting the validity of the regulative principle, but clearly Goldman would reject such a suggestion outright. At any rate, I believe there to be a distinction that provides a place for principles of both sorts and, further, which captures the difference between the internalist and externalist conceptions of justification.

I propose distinguishing between a person being justified in holding a belief and the belief itself being justified. What makes a person justified in holding a belief resides in the quality of his epistemic action. There is much that this can involve, including conduct of inquiry, evaluation of evidence, and consideration of plausible alternatives. In brief, however, let's just say that a person is justified in believing something to the extent that he holds the belief rationally and responsibly. Yet, a belief can be justified even in the absence of any action on the part of the believer, as in the case of beliefs formed automatically or routinely, without any deliberate consideration. Indeed, I suggest that most of our beliefs are of this sort, including run-of-the-mill perceptual, memory, and introspective beliefs. The distinction is clearest in the case of psychologically basic (noninferential) beliefs, since whatever would make someone justified in holding such a belief would also render that belief psychologically nonbasic. If a basic belief is justified at all, because it is basic there is nothing one does in order to be justified in holding it. Nothing counts as being rational or epistemically responsible in holding a basic belief. This is why, as we saw, Bonjour concedes that psychologically basic beliefs cannot satisfy the internalist conception of justification. If only he distinguished between people being justified in believing and beliefs being justified, he would not have to make this skeptical concession.

This distinction defuses an internalist argument offered by Kornblith. Insisting that “justified belief cannot be identified with reliably produced belief,” Kornblith argues,

Since epistemically responsible action may result in something less than reliably produced belief, an agent may be justified in holding a belief without that belief being reliably produced. Beliefs produced by unreliable processes, where the extent of the unreliability would not be detected by an epistemically responsible agent, are nonetheless justified.

In effect, Kornblith is equating a belief being justified with an agent being justified in holding it. If he distinguished the two, he could then say that an agent can be justified in holding an unreliably produced belief even if the belief itself is not justified. And, of course, he could say that a belief is justified even if it is not the case that the believer is justified in holding it (which does not mean that the believer would be justified in not holding it).

So I think that our distinction captures what is right about both conceptions of justification. They are conceptions of two different things! Internalism, taking rationality and responsibility to be the marks of justification, can maintain that whatever makes a person justified in holding a belief must be available to him. And externalism can maintain that being justified is whatever a true, ungettiered belief must be if it is to qualify as knowledge. As we have seen, a person can be justified in holding a belief even if it is not justified in that sense. For since the person could rationally and responsibly rely on some false generalization or principle, reliance on which cannot give him knowledge, he could be justified in holding a belief that is itself not justified.

Finally, our distinction undercuts Bonjour's seemingly decisive counterexample to externalism. He describes the case of a completely reliable clairvoyant who “possesses no evidence or reasons of any kind for or against the general possibility of such a cognitive power, or for or against the thesis that he possesses it,” and asks, “Is Norman epistemically justified in believing that the President is in New York City?” (p. 62) Now Norman either does or does not believe that he is a reliable clairvoyant, but by hypothesis he has no reason to believe this. So if he does, this belief is unjustified and thus cannot help justify his belief about the President's whereabouts. But if Norman has no belief about the reliability of his clairvoyance, his belief about the President's whereabouts is, according to Bonjour “epistemically irrational and irresponsible, and thereby unjustified,” since “part of one's epistemic duty is to reflect critically upon one's beliefs” (p. 63). Now even if fulfilling this “duty” is necessary for Norman to be justified in believing, that leaves open the question, once our distinction is drawn, of whether the belief is justified. The reliabilist can maintain that it is. Given the reliability of the process that leads to such beliefs, this process gives Norman knowledge whenever it results in a true belief (unless, of course, the belief has been gettiered).

Since he uses it to criticize Goldman, presumably Bonjour would take the example to show that simple reliability is not enough even for justified belief. Yet Goldman himself recognizes that simple reliability is not enough and requires in addition to the reliability of the process leading to the belief that there not be available a reliable defeating process. Thus Goldman could explain why Norman's belief is not justified, though the result of a reliable cognitive process, by the fact that Norman fails to reflect on the reliability of his clairvoyance. The process of doing so, which surely would incorporate inductive principles generally relied on by Norman, would lead him to doubt or even deny the reliability of his clairvoyance. After all, Bonjour has stipulated that Norman does not have inductive support for believing himself clairvoyant. If he did have such support, contrary to Bonjour's description of the case, his belief that he is reliably clairvoyant would be justified.

Indeed, perhaps simple reliability is enough, at least in the case of unreflective agents who, after all, are still capable of knowledge, hence of justified beliefs. Suppose that it does not occur to Norman to reflect on his powers of clairvoyance any more than it occurs to most people to reflect on their powers of perception. If perception can give them knowledge, hence justified beliefs, why couldn't clairvoyance do the same for Norman? It might seem that Norman is required to reflect on his clairvoyant powers in a way that ordinary people are not required to reflect on their perceptual powers, but what is the relevant difference? I think there is no relevant difference, and that there seems to be one only because of our doubts about clairvoyance in real life. Thus it seems that if Norman really is reliably clairvoyant and has no reason to believe otherwise, his beliefs based on that power are as justified as ordinary perceptual beliefs. To appreciate this imagine that what we take to be the perception of physical objects does not involve their affecting our sense organs. Instead, a benign Cartesian demon, recognizing the unbridgeable mind/body barrier, arranged the world so that our sensory experiences were generally veridical, just as they are (presumably) in fact. Then our knowledge of physical objects would be not by perception but by clairvoyance instead. We would all be in Norman's position, except that there would be no way to check the reliability of our clairvoyance. Indeed, in this hypothetical circumstance the situation with clairvoyance would just like our actual situation with perception, where simple reliability is enough for justified belief!

A conception of reasoning

Bonjour gives the impression that the dispute between internalism and externalism is solely about basic beliefs, but it goes further than that. This is evident from Goldman's formulation. His recursive definition of justified belief distinguishes categorically reliable from conditionally reliable belief-forming processes. Categorical reliability is defined for belief-independent processes, whcih lead to beliefs not based on other beliefs. Obviously these are what I have been calling psychologically basic beliefs. Conditional reliability is defined for belief-dependent processes, which lead to beliefs based (at least partly) on other beliefs. The reliability here is conditional since what comes out of the process depends on what goes into it. Now why doesn't Goldman content himself with reliabilism about basic beliefs and let the internalists have their way with inferential beliefs? As we have seen, Goldman thinks that internalism is relevant only to regulative principles of justification, not to descriptive, theoretical principles. Also, he is skeptical about the prospects of identifying and adequately formulating evaluative principles. Not only that, he points out that “doxastic habits” precede “the choice of a doxastic principle.” This is necessary, he argues, since otherwise “there would be an infinite regress of choices [of doxastic principles].” I agree with Goldman, but I think the issue goes even deeper than this. It concerns the very nature of reasoning.

I have been supposing (contrary to Bonjour) that most of our everyday, garden-variety beliefs qualify as knowledge, and that to do so they must be justified. Accordingly, a reasonable theory of justified belief (hence of knowledge) must take into account real-life limitations on our everyday reasoning. Now philosophers tend to focus on cases of deliberate, full-blown reasoning. Being interested in identifying the elements and the virtues of reasoning, they naturally turn to reasoning at its most explicit. Yet such reasoning is exceptional: most of the reasoning that gives us knowledge is largely inexplicit. It is what in Artificial Intelligence is called “default reasoning.”

The simplest kind of default reasoning occurs when, in considering a given issue, we believe the first thing that comes into our heads. This pervasive phenomenon can lead to justified beliefs insofar as it is reliable. Its reliability depends, as I will explain later, on how reliable we are at knowing when to think twice. Jumping to conclusions enables us to form beliefs much more freely than explicit consideration would allow. This does not mean that they are less justified (or are governed by lower standards of justifiedness), for in most cases explicit consideration would yield the same result--after considerable time and allotment of attention.

Moreover, even when our explicit reasoning involves many steps, at every step of the way we make implicit assumptions. We reason just as if we had made the assumption explicitly. That is, we reason in a way that is sensitive to the assumption, only making inferences consistent with it. Ordinarily we do not question such assumptions unless there occurs to us some reason to do. We rely on our reliability at detecting reasons for challenging our assumptions. Even if our lines of reasoning were always perspicuous, so that we could view them as a whole, there would still be points at which we do not check for validity but simply “go along” with the reasoning at that point. We just “see” that the next step follows. Thus, even when our reasoning is elaborate, to some extent it is still default reasoning.

Insofar as justified beliefs generally result from default reasoning, the internalist conception of justification, even when restricted to inferential beliefs, is psychologically unrealistic and epistemologically inadequate. Even if all aspects of a given piece of reasoning were subject to explicit evaluation, the reasoning does not have to be evaluated in every evaluable respect in order to lead to justified belief. It can include elements that are not explicitly evaluated, not to mention implicit assumptions that would become explicit if the step in question were explicitly evaluated. On the reliability conception, their implicit “evaluation” consists simply in their not being questioned. Such an evaluation is reliable insofar as one is reliable at detecting good reasons for questioning steps in one's reasoning. Typically, the believer is not aware of what validates the reasoning and might not even be able to be aware of all of it. Even if he were aware of what must be true for his reasoning to be valid, he might have no idea how to establish these underlying presuppositions. That would require, in effect, knowing how to answer the skeptic, and that's too much to ask of the ordinary cognizer (not to mention the seasoned epistemologist!).  Also, the default conception of ordinary reasoning suggests that what makes a belief justified is not merely the actual reasoning that leads to the belief. Would-be reasoning is relevant too, reasoning that would take place if thoughts of certain possibilities occurred to the person. This means that how justified the belief is depends on the reliability of the process of thinking of relevant possibilities (JRA's) and even on the reliability of the process whereby they are evaluated. As we shall see next, this is true even in the case of psychologically basic beliefs.

Taking Things for Granted

The default conception of everyday reasoning has an interesting application to Gilbert Harman's approach to the Gettier problem, which I think has been unjustly charged with being psychologically implausible. Harman proposes a psychologistic strategy of using intuitions about knowledge “to decide when reasoning has occurred and what reasoning there has been.” The strategy is based on the principle (P) that reasoning can give one knowledge only if it contains no false steps. Harman suggests that what distinguishes a given Gettier case from its normal counterpart is that the reasoning leading to the belief contains (essentially) something false. Now if principle P is to distinguish the two cases, the reasoning in the Gettier case must be elaborate enough to contain the requisite false step. But then the reasoning in the normal case, where there is genuine knowledge, has to contain a counterpart of that step. Thus Harman's strategy can seem psychologically unrealistic, in that it requires attributing implausibly elaborate reasoning to the normal believer. If I may allude to a few well-known examples, the believer seems not even to consider, much less affirm, that what he takes to be a barn is not a papier-mache facade, that the candle he seems to see directly he is not really seeing through a system of mirrors, or that Havit is not the student who owns a Ford. I suggest that the step required in the normal case corresponding to the false step in the Gettier example concerns a proposition which is not explicitly considered but is merely taken for granted.

Let us look at the second case, Harman's perceptual Gettier example. A person has a justified true belief that there is a candle in front of him. He does not know this, however, for although there is a candle in front of him, he is unaware (and has no reason to suppose) that what he is seeing is really the reflection in a mirror of a candle off to the side. However, in inferring that there is a candle in front of him, he does not seem to be explicitly supposing that what he is seeing is not the reflection in a mirror of a candle off to the side. Accordingly, Michael Williams argues that there is no evidence to warrant ascribing reasoning that does include this supposition. So how can Harman's strategy be defended against this charge of psychological implausibility?

Let us distinguish between reasoning realizing an inference pattern and its merely instantiating that pattern. A piece of reasoning realizes an abstract pattern of inference if it contains psychologically real elements corresponding to all the steps of that pattern. It merely instantiates that pattern if there is some step that is not explicitly included but merely implicitly assumed. This distinction makes sense, I suggest, if we suppose that our ordinary, routine reasoning, as in perceptual judgment, operates according to something like the following rule, which I call the taking-for-granted rule.

(TFG) If it seems to me that p, then infer that p, provided no reason to the contrary occurs to me.

If our routine reasoning relies on the TFG, this reliance leads to justified beliefs insofar as we are able to detect abnormal circumstances. I must be pretty good at knowing when not to infer that things are as they seem in order to be justified, when the situation is normal, in supposing that things are as they seem. If I were insensitive to abnormal situations, I would directly infer that p even when I should not. In following TFG, whenever I directly infer that things are as they seem, i.e. without considering reasons to the contrary, I implicitly rely on my reliability at detecting indications of abnormality.

I am suggesting that we jump to conclusions except when we look before we leap. That's obviously efficient, but how reliable is it? Offhand, jumping to conclusions seems to gain speed at the risk of error. It looks as though it could get us into lots of trouble. But don't forget, drawing inferences is, as Mill observed, “the only occupation in which the mind never ceases to be engaged.” We can't avoid trading off possible error for speed, for there are always more inferences to be made. If we didn't generally jump to conclusions, we wouldn't make most of the inferences that need to be made. In any case, it seems that when we do jump to conclusions, we are generally right. We are generally right in our snap judgments about the kinds and qualities of things we perceive around us, right in our recollections about prior experiences, right about persons, places, and things we seem to recognize, right about what people mean when they talk to us. Perceptual judgment, recall, recognition, and understanding utterances are all clear cases of generally reliable jumping to conclusions. Since this is not a monumental coincidence, somehow our inferences must take relevant information into account without getting bogged down in irrelevancies. But how? How do we resolve the tension between efficiency and reliability? After all, reliability requires ruling out alternatives to the tempting conclusion. The way this tension is resolved, I suggest, is that alternatives can be effectively and legitimately ruled out without even being considered, at least not consciously. This can occur if our reasoning processes have the following feature: we consider an alternative only when there is special reason to do so. Otherwise, without explictly thinking that the alternative does not obtain, we reason as if it does not.

Now obviously our reasoning can work like this only if we are equipped somehow to detect the presence of reasons for considering alternatives that we ordinarily take for granted not to obtain. A belief resulting from such a process is justified to the extent that the process not only leads to true beliefs, at least generally, but also guards against forming false beliefs by means of precautionary subroutines that are generally activated when and only when they need to be. For it is only to that extent that following TFG can lead to justified beliefs. It says, to repeat, “If it seems to me that p, then infer that p, provided no reason to the contrary occurs to me.” Accordingly,

TFG licenses me to jump to conclusions if I don't think of a reason not to. Thus, according to this model, the justification of such an inference is conditional on the nonoccurrence of a certain thought. In the case of visual belief, for example, ordinarily I assume that things are as they look, unless it occurs to me that my vision is being affected abnormally, say by bad lighting or devious psychologists. Similarly, in the case of recall, as of somebody's name or the spelling of a certain word, I take for granted that the first thought that comes to mind is the right one--unless it occurs to me that it might not be, say because some other possibility comes to mind.

If this picture of ordinary reasoning and its justification is at all correct, it has a fundamental consequence for the dispute between internalism and externalism. Since making inferences according to TFG requires the nonoccurrence of a certain thought, TFG has the remarkable feature that it cannot be explicitly followed. For it TFG occurred to me while I was following it, then I would have to consider whether there are occurring to me any thoughts to the contrary of my prospective conclusion, in which case I would no longer be drawing that conclusion directly. Instead, my reasoning would contain the additional thought that there are no reasons contrary to that conclusion. But that's not the way jumping to conclusions goes, or at least not the way it seems to go. I don't seem to draw my conclusion after noting that no contrary possibility has occurred to me, and if I did reason in that way, undoubtedly plenty of such possibilities would occur to me.

* * * * *

Only externalism is compatible with the supposition that our everyday reasoning is default reasoning and that this reasoning generally leads to justified beliefs and gives us knowledge. Besides, as argued earlier, internalism is much less plausible as a conception of justified belief than of being justified in holding a belief. I have not addressed the problem of precisely how to formulate externalism, specifically in its reliabilist form. Solving that problem would include finding a suitable way to individuate cognitive processes and specifying the precise role of back-up processes. It remains to be seen whether this can be done, for I have tried to show only that there is nothing wrong with reliabilism in principle.


[references omitted]