For classical internalism, the characteristically internalist nisus comes from deontology: warrant is conceived as epistemic duty fulfillment, and this, in the presence of plausible assumptions, leads directly to internalism. The Chisholm of Theory of Knowledge and Foundations of Knowing concurs; for the post-classical Chisholm, however, internalism persists but has lost its raison d'être. Coherentism and Bayesian coherentism are also forms of internalism, and are properly thought of as forms of post-classical Chisholmian internalism. Still further, however, there is also the interesting theory of John Pollock's “Epistemic Norms”1 and Contemporary Theories of Knowledge,2 a theory Pollock says is internalist. In this chapter I shall examine Pollock's conception of warrant. I propose to argue that Pollock is an internalist in name only; I shall also argue that his official view of warrant is deeply flawed; and I shall conclude that there are hints, in his thought, of a wholly different and much more promising conception.
I. Pollockian Epistemic Norms
Our question is, What is warrant or positive epistemic status? John Pollock offers a systematic and highly articulated answer-if not to that very question, then to one lurking in the nearby neighborhood. Suppose we begin by considering his account of justification and norms.
The important epistemological questions, says Pollock, bear on justification rather than knowledge:
Epistemology is ‘the theory of knowledge’ and would seem most naturally to have knowledge as its principle focus. But that is not entirely accurate. The theory of knowledge is an attempt to answer the question ‘How do you know?’, but this is a question about how one knows, and not about knowing per se. In asking how a person knows something we are typically asking for his grounds for believing it. We want to know what justifies him in holding his belief. Thus epistemology has traditionally focused on epistemic justification more than on knowledge. (p. 7)
So the central questions in epistemology, says Pollock, have to do with justification rather than knowledge; still, he does offer an analysis or account of the latter, and hence an account of warrant or positive epistemic status. To examine it, however, we must first consider his explanation of justification.
Neither the term ‘justification’ nor the notion of justification, of course, is unproblematic. To turn to the term, one might use it for whatever it is (enough of which) distinguishes knowledge from mere true belief. Pollock does not adopt this strategy, and (as I argued in chapter 1) there are excellent reasons for not doing so. First, it may be that what distinguishes mere true belief from knowledge is complex, something like the vector sum or product of two or more simpler qualities or quantities. Second, the term ‘justification’ has a deontological ring; it is redolent of rights and duties, permission and prohibition, blame and exoneration. As we saw in chapter 1, there is a long and impressive tradition-one going back at least to Descartes and Locke—according to which what distinguishes knowledge from true belief just is justification taken thus deontologically. But of course that is just one tradition among others; simply to baptize what distinguishes knowledge from true belief ‘justification’ is to give that tradition a confusing and undeserved advantage over its rivals.
So ‘justification’, as Pollock uses it, does not simply name whatever it is that epistemizes true belief: but then how does he use it? First, justification, he says, is essentially normative: “A justified belief is one that it is ‘epistemically permissible’ to hold. Epistemic justification is a normative notion. It pertains to what you should or should not believe” (p. 7). What is it that governs what it is permissible or impermissible to believe? Here we meet epistemic norms, the central characters in Pollock's epistemological drama:
Rules describing the circumstances under which it is epistemically permissible to hold beliefs are called ‘epistemic norms’. (p. 8)
Thus we can give an entirely adequate analysis of epistemic justification as follows:
A person's belief is justified if and only if he holds it in conformance to his epistemic norms. (p. 168)
So my beliefs are justified if and only if they are permitted by my epistemic norms. But what sort of flora or fauna are they: how do they arise and how do they work? What is the source of their normativity? Am I obliged to follow them? (The previous quotation suggests that if I don't, then I do what I should not do, what is in some way impermissible.) Suppose I don't appropriately conform to one of these norms: what sort of criticism is then appropriate to my condition?
1. The Nature of Norms
First, norms govern what Pollock calls ‘reasoning’: any change in belief, whether resulting from reasoning in the more narrow sense or not.3 (Perceptual and memory beliefs, therefore, will be acquired by reasoning.) Any belief can be evaluated epistemically; and any belief I hold will either be justified—epistemically correct or permissible-or else epistemically impermissible. (Not so for digestion: a given peristaltic contraction is neither permitted nor impermissible, not being subject to permission or prohibition at all.) His favorite example of an epistemic norm: “If something looks red to you and you have no reason for thinking it is not red, then you are permitted to believe it is red” (p. 169). He most often compares epistemic norms to what he thinks of as internalized norms for such activities as riding a bicycle, hitting a tennis ball, and typing-such norms as (for bicycle riding), “‘If you feel yourself losing momentum then push harder on the pedals’, and ‘If you think you are falling to the right then turn the handlebars to the right’” (p. 168). A crucial difference, as he sees it, between norms for activities of that sort and epistemic norms is that the former speak of what to do or what you ought to do, but the latter speak only of what you are permitted to do. A norm for bicycle riding tells you that you ought to do A under condition B; an epistemic norm tells you only that you may believe P under condition C.
So epistemic norms govern reasoning (broadly conceived) and are relevantly like internalized norms for bicycle riding and hitting a tennis ball. But how do they govern reasoning? Norms for some activities guide our behavior by way of our consciously holding them before the mind and explicitly conforming our actions to them. You follow a recipe for Mulligan Stew, or, more poignantly, a step-by-step set of directions for assembling a cardboard toy refrigerator for your daughter-the kind where the advertisement says “Takes no more than 20 minutes of your valuable time to assemble!” but in fact takes every evening for a week. This, says Pollock, isn't at all how things go with epistemic norms: we aren't typically aware of them or conscious of them; we don't bring them before our minds and consciously set out to form beliefs in accordance with them. We don't typically think: “now I am being appeared to redly; when one is appeared to redly (and all else is equal) it is permissible to believe that there is something red lurking in the neighborhood; so I'll believe that.”
But then how do epistemic norms guide our behavior? We already have the answer: they do so in the way in which internalized norms for bicycle riding or hitting a golf ball guide behavior. In such cases, when we first learn how to perform the activity in question, we begin by consciously following norms we explicitly think of or entertain. My cycling teacher says: “Remember: when falling to the right, turn the handlebars slightly to the right”; my driving instructor says “When you shift from first to second, let up on the accelerator pedal a split second before you depress the clutch; wait just an instant for the engine to slow down before you slip (not force) the gear lever into second.” I heed her words and at first consciously hold these directives before my mind. But soon such norms get internalized (however exactly that is to be understood); after some practice I can shift without thinking of the norms at all; after more practice the whole procedure becomes automatic. The same (says Pollock) goes for epistemic norms, except that typically we don't have to internalize them; they are, so to speak, internalized from the start. “The point here,” he says, “is that norms can govern your behavior without your having to think about them. The intellectualist model of the way norms guide behavior is almost always wrong.… Reasoning is more like riding a bicycle than it is like being in the navy [where they do things by the book]” (p. 129; Pollock's emphasis).
2. The Normativity of Epistemic Norms
But what makes just these norms-the ones that actually govern our belief acquisition and change, or the ones that should govern them-the right norms for reasoning? (Perhaps we should put it like this: of all the candidates for normhood, what is it that confers normhood on those candidates that are actually successful?) Pollock's characteristically bold answer: these norms are the right norms for reasoning because the norms we use are constitutive of the concepts we have. If, in doing a bit of reasoning, I had employed norms different from those I did employ, then I would not have done that bit of reasoning. More specifically, if, in forming a given belief B, I had employed norms different from the ones I did employ, then it would not have been B that I would have formed.
Let's see if we can come to a better understanding of this initially dark saying. Pollock begins by rejecting what he calls “the logical theory of concepts.” In essence, he says, this theory holds that what individuates a given concept, what makes it the concept it is, is its logical relations to other concepts (pp. 143–44). This idea generates the “picture of a ‘logical space’ of concepts, the identity of a concept being determined by its position in the space, and the latter being determined by its entailment relations to other concepts”4 (p. 143). So what constitutes a concept, what makes it the concept it is, on this theory, are its logical relations to other concepts. The concept being red, for example, entails the concept being colored (that is, it is not possible [in the broadly logical sense] that something exemplify the former but fail to exemplify the latter); it precludes the concept being a prime number (that is, it is not possible that there be an object that exemplifies both concepts); and it neither includes nor precludes the concept being square. Further, given any standard understanding of necessity, it is necessary that being red is related to these other concepts in those ways; it is necessary that it includes being colored, precludes being a prime number, and neither includes nor precludes being square. But of course a concept C precludes a concept C∗ if and only if it includes the complement—C∗ of C∗; it suffices, therefore, to think about the concepts C includes. So consider the conceptual train of being red—that is, consider the concepts it includes: it is not possible, on the logical theory, that being red should have failed to have that conceptual train. But it is also impossible, on the logical theory, that some concept distinct from being red should have had that very conceptual train. Having that conceptual train, therefore, is, on the logical theory, the (or an) individual essence of the concept being red.5 Indeed, it is something even stronger, since on this theory having that conceptual train is what makes being red the concept it is.
Now Pollock rejects this view—quite rightly, since it seems clearly false.6 In its place he proposes “the epistemological theory of concepts” (p. 147) according to which
concepts are individuated by their roles in reasoning. What makes a concept the concept that it is is the way we can use it in reasoning, and that is described by saying how it enters into various kinds of reasons, both conclusive and prima facie. Let us take the conceptual role of a concept to consist of (1) the reasons (conclusive or prima facie) for thinking that something exemplifies it or exemplifies its negation, and (2) the conclusions we can justifiably draw (conclusively or prima facie) from the fact that something exemplifies the concept or exemplifies the negation of the concept. My proposal is that concepts are individuated by their conceptual roles. The essence of a concept is to have the conceptual role that it does. (p. 147)
On Pollock's view, therefore, the essence of a concept is its conceptual role rather than, as on the classical view, its conceptual train. Of course, conceptual role and conceptual train significantly differ: one belief can be a prima facie reason for another even if it does not entail it; and one belief can entail another even if it is not a prima facie reason for it.7
But precisely how is this relevant in the present context? Our question was, What makes a given norm or norm candidate—for example, when appeared to redly, you may form the belief that you see something red if you have no reason to the contrary—a correct norm or a real norm? The answer: being appeared to redly is necessarily a reason (a prima facie reason) for judging that you are perceiving something red. This norm is constitutive of one or more of the concepts involved in my judgment that I see something red; and that means at the least that it is necessary, in the broadly logical sense, that it is a correct norm for that judgment. So the answer to the question, What makes this norm correct? is much like the answer to the question, What makes modus ponens valid? About all one can say is that it couldn't have been otherwise.
There is one further and fateful consequence Pollock draws from his epistemological theory of concepts:
Because concepts are individuated by their conceptual roles, it becomes impossible for people's epistemic norms to differ in a way that makes them conflict with one another. The epistemic norms a person employs in reasoning determine what concepts he is employing because they describe the conceptual roles of his concepts. If two people reason in accordance with different sets of epistemic norms, all that follows is that they are employing different concepts. Thus it is impossible for two people to employ different epistemic norms in connection with the same concept. (p. 148)
II. Justification and Objective Justification
Pollock's official account of epistemic justification goes as follows:
Epistemic justification consists of holding beliefs in conformance to correct epistemic norms. But as we have seen, our epistemic norms are constitutive of the concepts we have and hence it is a necessary truth that our actual epistemic norms are correct. Thus we can give an entirely adequate analysis of epistemic justification as follows:
A person's belief is justified if and only if he holds it in conformance to his epistemic norms. (p. 168).
Now clearly a person could be justified, in this sense, with respect to a true belief even if that belief does not constitute knowledge for him; justification, so construed, is not an apt candidate for warrant. For example, I might look at a red ball that is in fact but unbeknownst to me illuminated by red light, so that it would look red even if it were white. Then, says Pollock, I might be justified in believing the ball red but would not know that it was. Or consider such self-evident and nearly self-evident beliefs as 2 + 1 = 3, 7 + 5 = 12 and 4 is not prime. What would the norms governing such beliefs be like? The best candidates, I suppose, would refer to the phenomenology that goes with such belief. Two kinds of phenomenology accompany beliefs of this sort. First, there is something like a sort of broadly speaking sensuous phenomenology, a sort of phenomenal imagery (possibly variable from person to person). But second and more important, there is also a strongly felt inclination to believe the proposition in question; it is as if the belief in question has a sort of powerful perceived attractiveness. More accurately, there is a strongly felt inclination to believe that the proposition in question must be so, couldn't be false.8 Presumably norms for these beliefs would take the form of permitting such beliefs when they are accompanied by phenomenology of that kind.
Clearly, a belief could conform to that sort of norm when it had little or no warrant. Return to an example from chapter 3 (p. 59): I am captured by a group of Alpha Centaurian superscientists intent upon a cognitive experiment; in the course of conducting their experiment, they so modify my cognitive faculties that I believe every third proposition of the form n is prime(where n is a natural number between 23 and 200), disbelieving the rest. (Thus I believe that 23 and 24 are not prime, that 25 is prime, that 26 and 27 are not, and so on.) They further modify my faculties so that whenever I consider one of those propositions I believe—25 is prime, 28 is prime,…, 67 is prime,…, 199 is prime—I undergo the very sort of phenomenology that for me goes with simply seeing that, say, 5 is prime: I can apparently see that it is true in just the way I can see that 4 + 3 = 7 is true. Then in believing that proposition I am conforming to my norms for such belief; for they permit me to believe any proposition of this sort if it is accompanied by the right sort of phenomenology. So such propositions are justified for me, but they have little warrant for me. For warrant, says Pollock, we must turn to objective justification, which together with truth (and a codicil to accommodate “socially sensitive truths”) is sufficient for knowledge. The basic idea of objective justification is something like this. Suppose I am justified in believing P; I am also objectively justified in so doing, if, no matter what true propositions I came to believe, I would still be justified in believing P, and for the same reason that I am justified. We can put it a bit more formally as follows:
S is objectively justified in believing P if and only if: (1) S is subjectively justified in believing P and (2) There is a set X of truths such that, given any more inclusive set Y of truths, necessarily, if the truths in Y were added to S's beliefs (and their negations removed in those cases in which S disbelieves them)9 and S believed P for the same reason then he would still be (subjectively) justified in believing P. (p. 185)
This account, Pollock thinks, is insufficiently explicit; he therefore proposes a more complicated ‘official’ account of objective justification. The official account is of considerable interest (see Appendix); but the less explicit account will adequately serve our present purposes. Now “Objective epistemic justification,” says Pollock, “is very close to being the same thing as knowledge” (p. 185); all that remains to be added is a smallish qualification having to do with what (following Gilbert Harman) he sees as a social dimension of knowledge. “We are ‘socially expected’ to be aware of various things”: what is in our mail, what is announced on television, what any sixth grader has learned in school, and the like. A proposition is “‘socially sensitive for S’ if and only if it is of a sort S is expected to believe when true” (p. 192). Warrant, therefore, is given by objective justification plus a small qualification to take account of socially sensitive propositions.10
Pollock displays a certain degree of diffidence with respect to his account of knowledge: “At this stage in history it would be rash to be very confident of any analysis of knowledge” (p. 193). This diffidence, I think, is not entirely misplaced. I propose to argue that Pollock's account of warrant or positive epistemic status is seriously flawed, foundering on the possibilities of cognitive malfunction; I shall go on to argue that this account really represents a sort of transitional stage, an uncomfortable halfway house on the way to a more satisfactory view.11
A. Degrees of Warrant
First, a small problem I shall simply note without comment. Obviously warrant or positive epistemic status comes in degrees; obviously some of my beliefs have more by way of warrant than others. Pollock concurs (p. 5); yet his official account of objective justification makes no provision for degrees of justification, either objective or subjective. Pollockian norms are permissive (and possibly also prohibitive); and a person is justified in a belief “if and only if he holds it in conformance to his epistemic norms” (p. 168), that is, if and only if his holding that belief is permitted by those norms. But of course there aren't degrees of permission. You are permitted (or not) simpliciter: you can't be permitted to degree .3, say, or be more permitted to do one thing rather than another. (This points to a difficulty for any wholly deontological conception of positive epistemic status.) Similarly for objective justification. You are objectively justified in believing P, to put it crudely, if and only if (a) you are justified in believing P and (b) each truth is such that if you believed it, you would still be justified (for the same reason) in believing P; clearly then, there is no room for degrees of objective justification. It isn't at all easy to see how to handle this problem, but there are some possibilities,12 and sufficient ingenuity might do the trick.
B. Incorrect Norms and Identity of Concepts
I turn now to the deepest and most important difficulty with Pollock's account of warrant. It isn't easy to see precisely what the Pollockian norms are; but couldn't someone reason in accordance with mistaken or incorrect norms, so that even though her reasoning conforms to her norms, and even though she is objectively justified (again, according to her norms) nonetheless she fails to know what she believes? Return to Paul's plight in chapter 2 (p. 42). As the story went, Paul suffers from a brain lesion induced by radioactive fallout from a Soviet missile test. He now reasons differently from the rest of us; when appeared to in the church-bell fashion, he forms the belief that something is appearing to him in that fashion, and that it is orange. His noetic system is altered by the lesion in such a way that he now reasons according to the norm when you are appeared to in the church-bell fashion (and have no reason to think that what is appearing to you is not orange), it is permissible to believe that you are being appeared to by an object that is orange; he is appeared to in the familiar church-bell fashion and, in accordance with his norms, forms the belief that he is being appeared to by something that is orange. It isn't entirely easy to say when someone is objectively justified, but couldn't it be, in this case, that no matter what other relevant truths he had known, his norms would still have permitted him to form that belief in that fashion? Add that his belief—that he is being appeared to by something that is orange—is by happy coincidence in fact true: due to that cognitive malfunction he still wouldn't know; in fact he would be nowhere near knowing.
Or consider an epistemic agent S whose cognitive nature has been so altered—due to demon or black bile or perhaps to the “warmed or overweening brain” to which Locke attributed the enthusiasm of religious enthusiasts—that whenever he is appeared to redly, he finds himself with the belief that no one other than himself is ever appeared to redly. Objective justification, in Pollock's sense, is not an easy notion to work with; it can be difficult to determine what could be objectively justified and what not; but couldn't such a person be objectively justified in such beliefs? Add that his belief to this effect on a given occasion is true: perhaps there has been a nuclear holocaust; S's cognitive faculties are affected in the way suggested; the rest of us are so modified that we are no longer appeared to redly; due to trauma-induced amnesia, S is not aware of any of this. So when he forms his pathologically induced belief that no one else is appeared to redly, the belief he forms is true. Still, one wants to say, neither S nor Paul knows the relevant proposition; it is just by happy accident, here, that belief matches fact. And don't we have here an argument for the conclusion that the proposed necessary and sufficient condition for knowledge or warrant is not in fact sufficient?
Pollock's response, so far as I understand it, is that these examples, contrary to appearances, are not really possible. They are not possible, because different people cannot, contrary to appearances, employ appropriately different norms in reasoning:
Because concepts are individuated by their conceptual roles, it becomes impossible for people's epistemic norms to differ in a way that makes them conflict with one another. The epistemic norms a person employs in reasoning determine what concepts he is employing because they describe the conceptual roles of his concepts. If two people reason in accordance with different sets of epistemic norms, all that follows is that they are employing different concepts. Thus it is impossible for two people to employ different epistemic norms in connection with the same concepts. (p. 148)
In the above example, Paul has and uses a nonstandard norm for the concept x is appeared to by something that is orange—a norm that is not employed by the rest of us. This norm differs from ours in such a way that the reasoning it sanctions is not sanctioned by our norms. But then, says Pollock, it follows that Paul has not employed the same concepts as we. In particular he has not employed the concept x is appeared to by something that is orange; therefore he has not formed the belief I am being appeared to (in that church-bell fashion) by something that is orange. But then we don't have a case of someone's employing an incorrect norm and thereby failing to have knowledge: what we do have is someone's employing a correct norm for a concept distinct from x is appeared to by something that is orange—perhaps a concept the rest of us don't have at all.
Here we must be careful. Pollock does not hold that it is impossible to reason in a way out of accord with the correct norms for a concept; you can make a mistake, fail to reason in accord with your norms. You can fail to reason in accord with the correct norms for a concept; what you can't do is reason in accord with incorrect norms for a concept. What is not possible is that you have a norm for a concept that is not in fact a correct norm for that concept. Pollock's claim is that one can't reason in accord with incorrect norms.
But why not? What leads Pollock to this startling conclusion? How does the argument go? Recall that “the conceptual role of a concept” is “(1) the reasons (conclusive or prima facie) for thinking that something exemplifies it or exemplifies its negation, and (2) the conclusions we can justifiably draw (conclusively or prima facie) from the fact that something exemplifies the concept or exemplifies the negation of the concept”; Pollock's proposal is that “the essence of a concept is to have the conceptual role that it does” (p. 147); and what this means is that for any concept C and norm N, N is a norm for C if and only if it is necessary that N is a norm for C.
Suppose we concede for present purposes that this is correct.13 Let's agree first that there are norms for concepts; second, that these norms are typically permissive, as he says, so that they take such forms as In circumstances C you may believe that concept C∗ is exemplified; and third that if a given norm N is a norm for a concept C∗ then it is necessary that N is a norm for C∗. How does the conclusion follow; that is, how does it follow that if I form the belief that C∗ is exemplified but do not form it on the basis of a correct norm for C∗, then I haven't really formed the belief that C∗ is exemplified (so that it is not in fact possible to form that belief on the basis of an incorrect norm)?
It doesn't. Clearly there are incorrect norms for a concept (for example, Paul's and S's norms of a couple of paragraphs back) as well as correct norms. More exactly, it is possible to form beliefs involving a concept, employing the norms that are not correct norms for that concept. Clearly a person could form beliefs involving a concept on the basis of what are not in fact correct norms for that concept. Such a person would not, of course, be forming beliefs on the basis of the norms for the concept, that is, on the basis of the correct norms for the concept; but it does not follow that he wouldn't be forming beliefs involving that concept. If having such and such norms is an essence of a concept, then any statement falsely specifying that N is a norm for concept C will be necessarily false; but it does not follow that I can't reason in accordance with an incorrect norm—that is, a norm that is not in fact a norm for the concept in question. It is part of the essence (let's suppose) of the concept orange to have just the norms it does; but it does not follow that it is part of the essence of the belief this thing exemplifies orange to be formed only by someone whose norms for orange are those norms.14 Suppose one of Paul's norms for orange conflicts with the correct norms for that concept, so that he forms the belief orange is exemplified here on the basis of an incorrect norm for orange; it does not follow that this belief is not the belief, with respect to some object, that it exemplifies that concept. What follows is only that Paul's norms for orange are not the correct norms for it. If he does form that belief that way, then his reasoning is incorrect; but it does not follow that he can't form the belief that way. Here Pollock apparently fails to distinguish the claim that a given concept has essentially the norms it has (which follows from his theory of concepts) from the claim that a given belief has essentially the property of being such that it is formed by someone whose norms for the concepts it involves are correct norms (which does not follow from his theory).
Consider an ethical analogy. No doubt it is part of the essence of some (or all) actions to be permitted and enjoined in just the circumstances in which they are permitted and enjoined; there are correct norms for their performance. Suppose then that A is some such action and N1 − n the correct norms for it; it will then be part of the essence of A that N1− n are its correct norms. But it does not follow that being performed by someone who has N1− n for A is part of the essence of A, so that A cannot be performed by anyone who does not have N1− n as her norms for A. Perhaps it is part of the essence of such an action as causing someone severe pain that it is always wrong to do it just for the fun of it; it does not follow that I can't myself have norms for that action that permit doing it just for the fun of it. I would be acting wrongly; but I would still be performing the action causing someone else severe pain. In the same way, if Paul has incorrect norms for orange, it does not follow that he can't form beliefs of the sort that thing is orange; all that follows is that he is reasoning incorrectly.
So the argument begins from a dubious premise and is in any event inconclusive. What the argument really requires is a premise entailing that it is part of the essence of a concept, not just to have the norms N it does have, but to be such that no one can have other norms for it. The argument is unsound, therefore; and isn't the conclusion of the argument—that two people can't have or employ conflicting norms with respect to the same concept—wholly implausible? Consider the Pyrrhonian skeptic. Sextus Empiricus recommends that “in the hope of attaining quietude,” we withhold ordinary perceptual beliefs (among others), not acceding to our natural tendencies to form such beliefs when appeared to in the familiar ways.15 Thus “we are brought firstly to a state of mental suspense and next to a state of ‘unperturbness’ or quietude” (chapter 4). Sextus urges us to adopt different norms: his recommendation is that we replace such norms as
when you are appeared to redly and consider whether you perceive something red and have no reason to think that you do not, you may form the belief I am appeared to by something red
by such norms as
when you are appeared to redly and consider whether you perceive something red and have no reason to think that you do not, you may withhold the belief I am appeared to by something red
or, more radically, by the norm
when you are appeared to redly, you should not form the belief I am appeared to by something red.16
As Reid and Hume point out, it is at the least very difficult to withhold ordinary perceptual beliefs under the conditions in which, in the course of ordinary life, we form them. Still, given sufficient effort and training, perhaps a talented person could do it; perhaps Sextus himself managed to govern his belief in the way he recommended. At any rate it clearly seems possible (in the broadly logical sense) that he do so. Of course, if he did manage this feat, his formation of such beliefs as I see something red would be governed by norms different from the one that governs our formation of such beliefs—different, and conflicting in the relevant way. A belief in accord with his norms would not be in accord with ours. But then on Pollock's view, Sextus wouldn't so much as have the concept x sees something red; hence he could not form the belief in question. So (on Pollock's view) it is not just unwise or ill-advised to follow Sextus's recommendation; it is logically impossible. Indeed, if Sextus managed to follow his own advice, he would no longer be able to conceive that advice. For if he followed it, he would no longer be able to form the belief I see something red and a fortiori could form no belief of the sort when appeared to redly, it would be good to withhold the belief I see something red. But surely this is much too strong. It may be difficult to follow the instructions in question; it may be foolish to do so; but it is not logically impossible. Surely there could be or could have been beings who formed beliefs in accordance with the norms Sextus suggests.
For another example, consider John Locke, who deplored the ways in which the generality of mankind forms belief. There is one “wrong measure of probability,” he says, that “keeps in ignorance, or errour, more people than all the others together”; this is “the giving up our assent to the common received opinions, either of our friends, or party, neighbourhood, or country”17 His idea is that the unreflective run of mankind form their beliefs in accord with what Reid calls “The Principle of Credulity”; when one of our peers tells us something, we tend to believe it, at least in the absence of countervailing reasons. (Reid plausibly thinks this belief-forming principle is native to us.)18 So the generality of mankind reasons in accord with such norms as if practically everyone I know believes p, then I may (or ought to) believe p; Locke recommends that we eschew norms of that sort and try to form our beliefs on the basis of such quite different (and conflicting) norms as I may (or ought to) believe p only if that proposition is sufficiently probable with respect to my evidence. But then on Pollock's view Locke's advice has nothing whatever to be said for it. For if in fact some of us do form the belief in question in accordance with the first norm but not the second, then it is not even possible that someone should form beliefs of that sort in accord with the second but not the first. So the advice is either logically impossible to follow or else such that everyone already follows it.
Once more, this is surely incorrect. Locke's recommendation is not defective in that way, even if it is unrealistic (perhaps even Quixotic and foolish). Clearly there could be people who formed beliefs in accord with the first norm but not the second, and people who formed beliefs in accord with the second but not the first. On the face of it, people can form beliefs in accordance with all sorts of conflicting norms. The paranoid forms the belief that Fred is out to get him and seems to do so in accord with norms the rest of us do not employ (“that subtle, slightly squinty-eyed look gives him away”): but isn't it the belief Fred is out to get him that he forms? If not, why think he is paranoid? And isn't it possible (even if improper) to form one's beliefs according to the norm if believing p is in your best interests (or would afford you pleasure or comfort) then you may believe p?
Second, recall that there must be norms, not only of the form when in circumstances C you may believe P but also when in circumstances C you may believe P to degree d. Or rather, recall that norms will typically be of the latter sort, not the former; what one does, when one forms a belief, is to form a belief of a certain strength, and obviously not just any degree of strength is appropriate. (You are driving through Chicago; you catch a quick glimpse of what appears to be a hippopotamus in the median strip; if you form the belief that's a hippopotamus and hold it as firmly as you believe that Chicago is in Illinois, then your degree of belief is inappropriate.) Now even if it is psychologically impossible to withhold ordinary perceptual beliefs, it certainly seems possible to hold these beliefs with a somewhat different degree of firmness than is customary. Couldn't Sextus adopt and come to exemplify norms licensing degrees of belief somewhat different from those of the rest of us? Perhaps he holds pretty much the same perceptual beliefs as the rest of us, but holds them more tentatively. But again, on Pollock's suggestion, even this would be impossible; for, on his suggestion, if the skeptic's norms are different from ours—if (under given circumstances) they license forming a belief with a degree of strength different from what our norms license—then the skeptic wouldn't so much as have the relevant concepts.
Still further, our norms are modified or shaped by what we learn. We learn to trust some people under some conditions, but distrust others under others. We learn to exercise a certain skepticism about what we are told by politicians; we learn to take more seriously what someone tells us when what she says does not redound to her own credit (unless she is the sort of person whose apparent self-deprecation is a subtle form of self-aggrandizement); we learn not to form beliefs about marital discord without speaking to both parties; and so on. But then our norms are also thus shaped and altered by what we learn. It would be monumentally implausible to suppose that whenever my norms alter under the pressure of experience, I lose a concept I previously had, replacing it with a new one.19
The same goes for a wide variety of cases. You and I both have the concept red; we therefore exemplify norms for the formation of beliefs of the sort I see something red or that thing is red. For a given set of circumstances, there will be an interval d—d∗ such that my norms license the formation of a belief of that sort to any degree within that interval. But surely there is no reason to think it is necessary, in order for us both to have the concept red, that our norms coincide on these intervals; why couldn't they diverge, even if only slightly? Hume points out (and this may have been known even before Hume) that we reason inductively; we form beliefs reflecting a sort of expectation that the future will be like the past in relevant but hard to specify respects. So there are norms licensing my belief that the sun will rise tomorrow (given my past experience); these norms, no doubt, will involve intervals of degrees of belief. Perhaps your norms specify slightly different intervals from mine: must we conclude that we do not have the same concepts? Similar remarks obviously hold for memory, and for a priori beliefs. You and I both believe, say, that there neither are nor could be nonexistent objects; but I believe it more firmly than you. Couldn't this reflect a difference in the norms governing our belief here? For a given phenomenology of apparent obviousness, say, couldn't my norms specify a slightly different interval from yours? It is surely hard to see why not, and hard in excelsis to see that this is logically impossible.
Still further: clearly there could have been creatures capable of the same beliefs as we, but with different cognitive powers; and we ourselves could have been differently constituted. Take the concept orange, for example; and suppose the world had been different in such a way that there is always something orange near anything that is blue. Surely God could have so created us that among our norms would have been such items as when you are appeared to bluely, you may form the belief that there is something orange nearby—even though that norm is not in fact a correct norm for orange. Or (to take an example that involves what Pollock calls an attribute rather than a concept) we could have been like pigeons, having a built-in way of telling directions. Then we could be taken blindfolded through a complicated maze, spun around repeatedly, and still been able to tell which way is south. If we had been so constructed, we could have had norms for south that are different from the ones we do have—norms involving a sort of phenomenology that none of us has ever experienced. But we should still have been able to form such beliefs as Mexico City is south of Minneapolis. And here we don't just have variation, within the relevant norms, with respect to degree of belief sanctioned, but with respect to the formation of the belief itself.
Once it is clear that we can share a concept even though we differ in the above fashions with respect to the norms we adopt, it is hard to see why we can't differ even more radically. To return to the examples with which I began: doesn't it seem possible that (due to pathology) Paul should begin to form his beliefs quite differently, exemplifying norms different from those he exemplified before, exemplifying such norms as when you are appeared to in the church-bell fashion, you may form the belief I am appeared to by something that is orange? Doesn't it seem possible that (due to psychic disorder) S could come to exemplify such a norm as when you are appeared to redly, you may form the belief no one else is appeared to redly?
Pollock claims that what are in fact the correct norms for a given concept are constitutive of that concept, where this means at any rate that it is necessary that those norms are correct for that concept: from this he infers that it is not possible that different people form the same belief while employing or exemplifying conflicting norms. What we have seen, however, is that the argument is invalid, and the conclusion implausible. But then Pollock's view as to what it is that distinguishes mere true belief from knowledge—that is, being objectively justified—is incorrect; an unfortunate of the sort just described could be objectively justified in his belief despite its having little or no warrant or positive epistemic status for him.
IV. New Directions
I believe that Pollock's position is an uncomfortable way station on the journey from the traditional and explicitly deontological conception of warrant in his first book Knowledge and Justification to a wholly different sort of view. Suppose we begin by noting a certain problem with his account of norms. He suggests that on the “intellectualist model” the picture as to how norms direct our reasoning is too intellectualistic; but doesn't the same hold for his own official view? He compares the process of belief formation, retention, and change to such activities as bicycle riding, typing, hitting a golf ball, and so on (p. 171), and argues that epistemic norms govern belief in the way norms for those activities govern them. There are analogies here: each is something one comes or can come to learn to do; each requires a certain maturing process; each can be done without paying much conscious attention to the process, and so on. But there are also significant differences. In the first place, every well-formed human being learns to reason, in Pollock's broad sense (and learns to do so by an early age); not so, of course, for typing and cycling. In this respect, reasoning is more like walking and running. But it differs significantly even from them: it is not typically under direct and conscious control. If I want something from the refrigerator, it is up to me whether or not to walk to it (as opposed to crawling or running to it, or asking you to go to it for me). I can refrain from walking and running; I can give them up for Lent if I choose. But I can't give up reasoning for Lent—not, at least, without giving up a great deal more.
Furthermore, I can undertake to walk or run; and I can decide, if I choose, how long to make each stride, and which direction to walk in. I can also decide to walk backward, or in some weird, ridiculous way.20 Pollock sometimes seems to suggest that the same goes for reasoning. He sometimes speaks as if he thinks we typically undertake to form beliefs: “The sense in which the norms guide our behavior in doing X is that the norms describe the way in which, once we have learned how to do X, our behavior is automatically channeled in undertaking to do X” (p. 131); and he sometimes speaks as if he thinks we typically deliberate about what to believe, the epistemic norms guiding us in coming to a decision as to what to believe: “I have taken the fundamental problem of epistemology to be that of deciding what to believe.… Considerations of epistemic justification guide us in determining what to believe” (p. 10).
But the fact is in typical cases I neither undertake to believe (anymore than I undertake to breathe) nor make any decisions as to what to believe. I have too little direct control over my beliefs for that. I consider the corresponding conditional of modus ponens: I find myself with an ineluctable inclination to believe that this proposition is true and indeed necessarily so. You ask me what I had for breakfast: I find myself believing that what I had for breakfast was a grapefruit. I am appeared to redly; I find myself with the belief that I am perceiving something red. I consider the question what Caesar had for breakfast the morning he crossed the Rubicon: I find myself with no belief on that topic. In each of these cases (as in general), I have little or no direct or conscious control. I can't just decide to accept, say, Affirming the Consequent (or Ignoring the Antecedent) instead of modus ponens; I can't just decide not to believe that I had a grapefruit for breakfast; and I can't just decide to form a belief as to what Caesar had for breakfast that fateful morning (although I can decide to go to the library and look it up). Or rather, whether or not I can decide to do these things, I can't in fact do them. As Thomas Reid says, “My belief is carried along by perception, as irresistibly as my body by the Earth. And the greatest sceptic will find himself to be in the same condition. He may struggle hard to disbelieve the informations of his senses, as a man does to swim against the torrent; but ah: it is in vain.”
This is not to say, of course, that I have no control at all over my beliefs. I can arrange that I will have beliefs on certain topics by putting myself in the right conditions; I may also resolve to be less credulous (or less skeptical), pay more attention to the evidence, not insist on evidence when it is not appropriate, fight my tendency to form beliefs by wishful thinking, and the like. Nor am I saying that a person couldn't learn to inhibit her natural belief-forming tendencies. Perhaps I could follow Sextus's advice and (by dint of long and arduous training) attain a state in which I do not believe that I see something red when I am appeared to redly. (Perhaps I could even train myself to believe that I am appeared to by something green under those conditions.) But the point is that if these things are humanly possible at all, they are very difficult: for the most part we have little direct control over what we believe, and neither undertake nor decide to form specific beliefs.
Pollock compares epistemic norms most explicitly to norms we first consciously grasp and follow in a step-by-step explicit fashion and then internalize: “[Epistemic norms] describe an internalized pattern of behavior that we automatically follow in reasoning, in the same way we automatically follow a pattern in bicycle riding. This is what epistemic norms are. They are the internalized norms that govern our reasoning” (p. 131). But for the most fundamental kinds of beliefs, there is typically nothing like internalizing our epistemic norms. Under the right conditions, I believe that I am appeared to redly, or that there is a tree outside, or that 7 + 5 = 12, or that you are coming for dinner. Perhaps, in a semi-Pickwickian way, it is correct to say that I know how to believe these things and have learned how to do it; but it isn't as if there is anything like internalizing a set of directions for forming such beliefs. We don't first learn how to believe such things by thinking about maxims or norms for beliefs, explicitly following them by noting that we are in the condition in which, according to the norm, it is permissible to form the belief in question, and then forming that belief. Norms for bicycling or driving get internalized and become second nature; most epistemic norms, by contrast, are first nature. They don't have to become internalized; they regulate our doxastic carryings on long before we so much as notice them. Epistemic norms are more like the ‘norms’ that govern perspiration, or adrenaline flow, or blood pressure, or rate of respiration. To think about them on the pattern of internalized rules or directions is to think about them in too intellectualistic a way.
In a remarkable section of Contemporary Theories of Knowledge, Pollock himself suggests a wholly different way of thinking about norms. He remarks that “a fuller understanding of the nature of epistemic norms can be obtained by seeing how they are integrated into the broader picture of man as a cognitive machine” (p. 149); he suggests that “considerable light can be thrown on human epistemology by reflecting on the working of cognitive machines in general”; he considers how we might try to build an information processing machine (‘Oscar’) that simulates our cognitive behavior. In such a machine, he says,
Sensory input results in behavioral output, and an important part of the connection is provided by thought. The thought processes constitute reasoning and are governed by rules for reasoning.… The rules for pure reasoning constitute epistemic norms. In effect, epistemic norms comprise a “program” for the manipulation of sentences in the language of thought in response to sensory input. (p. 161; emphasis added)
Since Pollock sees a belief as something like a sentence in the language of thought (a sentence treated in a certain special way), his view is that epistemic norms relevantly resemble a program for an information-processing machine. Such a program describes how the machine will behave under various conditions (if it is functioning properly). Figuratively speaking, it instructs the machine to take certain lines of action in response to a certain state of affairs (when you are in state S and condition C obtains, move to state S∗); among other things it specifies the conditions under which the machine in question will ‘believe’ certain propositions. To turn to human beings and remove the quotation marks, epistemic norms will be very much like descriptions of the doxastic behavior of a properly functioning human being. When appeared to in a certain characteristic way, I will form a certain characteristic belief (if there are no defeaters); when I consider an instance of modus ponens, it seems obviously valid and I form the belief that indeed it is; when you tell me that your name is ‘Alexander Hamilton’, I believe you. Epistemic norms will be like generalizations of these descriptions.
But note how very different this conception of norms is from the one to be found in Chisholm or, indeed, the entire Lockean—Cartesian tradition—the tradition in which the notion of justification has its natural home. When Oscar fails to function in accord with her specifications, we may rightly think she is defective, or isn't working properly; but we can hardly claim that she isn't justified in functioning that way. (She should be ashamed of herself?) The notions of duty, obligation, permission, exoneration, blameworthiness, justification—that whole deontological stable—these seem irrelevant to Oscar and her functioning. Here we have a conception of epistemic norm that differs toto caelo from the deontological conception. Norms thought of this new way are more like specifications for a piece of machinery. Thus the specifications for a 1985 GMC van say something like (I'm just guessing) “After a cold start, the engine idles at 1,500 RPM until the coolant temperature reaches 180°F; then it slows to 750 RPM.” Such specifications describe how a machine of a certain sort functions when it is working properly.
So epistemic norms in this new conception are like specifications for a mechanism, or descriptions of how a certain kind of device functions when it is working properly. Perhaps you think it is stretching things to call specifications or descriptions ‘norms’. But the fact is that term is correct (if analogical); and there is a sense of ‘ought’ to go with it. Referring to the engine, we may correctly say “It ought to slow to 750 RPM when it heats up”; referring to the engine's thermostat we may say “It ought to open when the coolant temperature hits 180°F”; referring to your newly purchased but recalcitrant word-processing program you may say “When you strike the option and Q keys, it ought to align the right margin.” (This use of ‘ought’ is not predictive; we can say the same sort of thing even if we are constructing a new machine that has not worked properly so far and probably won't for the next month.) So such descriptions or specifications can rightly (if analogically) be called ‘norms’; they specify the behavior of a normal, properly functioning human person, that is, one whose functioning conforms to the relevant norms. In Warrant and Proper Function I argue that a crucial element of what confers warrant upon a belief for me is its being produced in me by my faculties functioning properly in a congenial epistemic environment. Equivalently, I argue that a belief has warrant for me when it is produced by my cognitive faculties functioning in accord with my design plan in an appropriate epistemic environment (where the element of the design plan governing its production is successfully aimed at truth). Pollock's deep view of the nature of epistemic norms is very much in accord with that suggestion; his norms, so thought of, begin to look very much like elements of the design plan.
There is another way to see that the conception of norm here adumbrated is wholly different from that to be found in the deontological tradition (the tradition of Pollock's earlier epistemological work and the tradition where the notions of justification and permission are at home). Consider the sense in which Pollock's new view is internalist:
It is easy to see that they [that is, epistemic norms] must be internalist norms. This is because when we learn how to do something we acquire a set of norms for doing it and these norms are internalized in a way enabling our central nervous system to follow them in an automatic way.… In general, the circumstance-types to which our norms appeal in telling us to do something in circumstances of those types must be directly accessible to our automatic processing systems. The sense in which they must be directly accessible is that our automatic processing system must be able to access them without our first having to make a judgment about whether we are in circumstances of that type. We must have non-epistemic access. (p. 133)
In what way, then, is Pollock's theory internalist? A person is justified in a belief if and only if the belief is formed in accord with her norms; such a norm must specify conditions under which it is permissible to form a certain belief; and the norm is internalist in the sense those conditions must be “directly accessible to our automatic processing systems”—that is, accessible to those systems without our having to make any judgments. So those conditions must be accessible to our central processors, whatever those are, in approximately the way in which the ambient temperature is accessible to a thermostat. The thermostat embodies such norms as when it gets below 70°F., close the switch. But then of course the ambient temperature (or something appropriately connected with it) must be ‘accessible’ to the thermostat.
So Pollock's theory is internalist in that on his theory there are norms that are internalist in the sense just outlined. But that sense is at best attenuated (not to say eviscerated and emasculated). It does not require, of course, that S is or even could become aware, either of the norms or of those conditions; the conditions have to be accessible to S's automatic processors, but of course they don't have to be accessible to S. The ph level of my blood is too low; my body makes the appropriate response; so the ph level of my blood must be accessible to my “automatic processors.” Does my believing this justify me in claiming to hold an internalist theory of acidulous behavior? You might concede that under certain conditions accessible to your “automatic processors,” your gallbladder will pump bile into your stomach. Would it follow that you endorse an internalist theory of bilious behavior?
And couldn't even the most blatant externalist be an internalist in this sense? You are an externalist; you think that what confers warrant upon a belief is its being produced by a reliable belief-producing mechanism. But you could consistently add that there are Pollockian norms governing the formation and sustenance of belief, and that conditions specified in the norms are accessible to the relevant “automatic processing system.” That wouldn't make you much of an internalist. Again, note how far this alleged sort of internalism is from that implied by the classical deontological conceptions of warrant. There the idea was that we ourselves can consciously regulate our beliefs, and are obliged to do so in a certain way. According to Locke, for example, I am blameworthy if I accept a belief B when it does not (after reflection) seem to be probable with respect to what is certain for me. But then of course I must be aware of or know or be able to know whether I accept B, and what is certain for me, and whether (after reflection) 6 seems more probable than not with respect to what is certain for me. Nothing at all like this is involved in Pollock's alleged internalism.
There is a strain, a tension, a schism in Pollock's thinking about norms. He thinks about them partly in ways appropriate to a deontological conception of warrant, as if they specified conditional duties or obligations: he speaks of permission, justification, being within one's rights, and the rest of the deontological panoply. But he also thinks of them as if they were more like directions embodied by a piece of machinery, or specifications of how an organism works when it is functioning properly—part of the design plan that is featured in the account of warrant I give in Warrant and Proper Function; and this is to think of them (and of warrant) very differently indeed. This latter way of thinking of warrant seems to me to have great promise. In Warrant and Proper Function I shall try to develop it in detail.