We have been examining the conception of warrant I call ‘classical Chisholmian internalism’: the conception developed in Chisholm's work up through Theory of Knowledge (TK) and Foundations of Knowing (FK). This conception is classically Chisholmian, but also classically internalist, in that it displays the motifs of the classical epistemic internalism of Descartes and Locke—an internalism that arises from deontology, from considerations of duty and obligation. Classical Chisholmian internalism displays a certain interesting development. Consider those epistemic principles to be found in TK and FK: noncontingent conditionals whose antecedents specify that S is in some condition or other and whose consequents specify that some belief of S's has some degree or other of warrant or positive epistemic status. Now in the first (1966) edition of TK, interestingly enough, these principles1 were all so stated that their antecedents, if true, specify that S is in a self-presenting condition: he is being appeared to redly, or believes that Albuquerque is in New Mexico, or is trying to get to Boston.
These self-presenting conditions, Chisholm thinks, are the ground of the warrant enjoyed by a belief B; they are what confers such warrant upon it; they are (if you like supervenience) that upon which warrant supervenes. But they also display another interesting peculiarity: it is also plausible to think that one cannot mistakenly believe that she is in such a condition. The mark of a self-presenting condition (FK, p. 10) is that necessarily, if a person is in the condition and considers whether she is, she believes that she is: hence it is impossible, when you are in such a condition, to consider whether you are and fail to believe that you are. It is equally plausible to suppose, however, that it is impossible, when you are not in the condition in question, to consider whether you are in it, and believe that you are. A self-presenting condition has this distinction: you cannot be mistaken in thoughtfully or reflectively attributing it (or its complement) to yourself. So self-presenting conditions are epistemically internalist conditions: conditions to which the agent has some special epistemic access. Indeed, they are internalist in excelsis, in that the agent enjoys a sort of infallibility with respect to them.
Examination of our principles makes clear that, according to them, our knowledge is not a function merely of what is self-presenting. Principle P2 refers to what I have called “the uncontradicted”; this involves the logical relations that one attribution may bear to others. If these relations obtain, that they obtain will not be self-presenting. (p. 25)
The internalist requirement, however, though not satisfied as straightforwardly, is nonetheless still satisfied. Chisholm continues:
But, I would say, one can always ascertain by reflection whether or not they obtain. Similar observations hold of “the epistemically unsuspect” (that which is “epistemically in the clear”), referred to in principles P3, P4 and P5. (pp. 25–26)
So principles P3, P4, and P5 are not such that their antecedents specify self-presenting conditions. Nevertheless, the conditions they do specify are still of a sort to which we have special access: their presence can be determined “by reflection alone.” The last principle, however—Principle P6—differs from the other five in that whether its antecedent holds is not something S can determine by reflection alone. For this reason, significantly enough, Chisholm proposes to call P6 not an “epistemic” principle, but a “quasi-epistemic” principle.4
So there is a certain movement away from the strongest forms of epistemic internalism. In FK the conditions specified in the antecedents of the principles are no longer self-presenting. Nevertheless, those conditions are still such that we have special access to them. They may not be self-presenting for us—our access isn't as special as all that—but we can still always determine by reflection alone whether or not they hold. And the source of this epistemic internalism, as we have seen, is epistemic deontologism, the idea that warrant is essentially a matter of aptness for epistemic duty fulfillment.
I. Post-Classicalism Explained
But a significantly different understanding of warrant appears in some of Chisholm's more recent work. In “The Place of Epistemic Justification”5 Chisholm takes the locution ‘x is more reasonable for S at t than y’ as primitive, partially explaining it by laying certain constraints upon it.6 What is most striking, however, is that here Chisholm no longer explains reasonability (and hence warrant) in terms of a duty to try to achieve epistemic excellence. As a matter of fact he proceeds in precisely the opposite direction: “I have previously written incautiously, that one's primary intellectual duties are to acquire truth and to avoid error. What I should have said is that one's primary intellectual duties are to believe reasonably and to avoid believing unreasonably.”7 This is a considerable step away from the views we have been considering. Previously, warrant was explained by way of the satisfaction of epistemic or intellectual duties; reasonability (and hence warrant) was to be understood in terms of the fulfillment (or aptness for fulfillment) of the general duty or obligation to try to achieve epistemic excellence. Here, on the contrary, warrant is to be explained in terms of believing reasonably, and one's intellectual or epistemic duties are to believe reasonably and avoid believing unreasonably. But then of course we cannot also understand a belief's being reasonable (and hence having warrant) for S as a matter of its aptness for the fulfillment of S's epistemic duty.
When we turn to Chisholm's “Self-Profile,”8 we find a deeper development of the same notions. There is, first, reiteration of the claim that epistemic concepts are normative concepts; but once again, there is a move away from the idea that they are deontological concepts. More significant, however, there is here the Brentanoesque suggestion that these normative epistemic concepts pertain not to duty and permission but instead to intrinsic value. “What I will say presupposes Aristotle's insight according to which knowing is, as such, intrinsically good.” On classical Chisholmian internalism, warrant is a deontological property; on the the post-classical view, however, it is an axiological property; it pertains to value and goodness rather than to duty.
To understand the post-classical Chisholm we need the notion of a purely psychological property. Although he gives a chain of definitions culminating in a definition of this notion,9 what he says about it more informally (together with a couple of examples) may suffice: purely psychological properties, he says, are “those properties to which we have privileged access. Every such property is necessarily such that, if a person has it and if he attributes it to himself, then his attribution is evident in the strongest sense of the term.”10 Examples, I take it, would be being appeared to redly, judging that 7 + 5 = 12, trying to achieve epistemic excellence, hoping that you will be able to repay the money you unwisely borrowed from your children, and the like. Chisholm then explains that a person's “evidence-base” at a time is “the conjunction of all the purely psychological properties that that person has at that time”; he replaces “A is more reasonable for S than B” by “A is epistemically preferable to B for S” and continues:
Generally speaking, we may say that, if taking a doxastic attitude A is epistemically preferable for a person S to taking a doxastic attitude B, then S's evidence-base is such that having that evidence-base and taking A is intrinsically preferable to having that evidence-base and taking B.
He goes on to say that we can systematically reduce our epistemic concepts to those of the theory of intrinsic value, an example of such a reduction being
(ED8) Believing p is epistemically preferable for S to believing q = def. Those of S's purely psychological properties which do not include believing p and believing q are necessarily such that having those properties and believing p is intrinsically preferable to having those properties and believing q.11
Clearly this suggestion fits with that of “The Place of Epistemic Justification,” according to which our epistemic duty is to believe reasonably. The state of affairs consisting in my believing reasonably is an intrinsically good state of affairs; furthermore, it is my epistemic duty to pursue it, to try to realize or actualize it. But then of course believing reasonably is not itself to be understood by way of the fulfillment of epistemic duty. Consider a case in which someone S reasonably holds some belief B: on classical Chisholmian internalism what makes B reasonable for S is that by accepting it S can fulfill her epistemic duty to try to achieve epistemic excellence. On the post-classical, axiological view, however, what makes B reasonable for S is just that S displays a certain evidence-base E, where B and E are so related that the state of affairs consisting in someone's having E and holding B is an intrinsically good state of affairs. This post-classical conception is clearly internalist, in that warrant or positive epistemic status is explained in terms of purely psychological properties to which the epistemic agent has privileged access. The internalism, however, does not arise from deontology; for it is not the case that warrant is to be understood in terms of such concepts as aptness for the fulfillment of epistemic duty.
According to the classical, deontological Chisholm,12 what confers warrant is a matter of doing one's epistemic duty, acting in an epistemically responsible manner, doing that which is such that neglecting it is blameworthy. According to the post-classical axiological Chisholm, on the other hand, what confers warrant is there being the right relationship between evidence-base and belief, a relationship such that when they stand in it, then the resulting whole displays a certain intrinsic value. The sort of intrinsic value involved, as we shall see in more detail, is not specified. Depending on the sort involved, we can see several different contemporary views as versions of post-classical Chisholmian internalism: coherentism, for example, in both its classical (see the next two chapters) and Bayesian (see chapters 6 and 7) manifestations, as well as the evidentialism noted in chapter 1 and the different versions of evidentialism to be noted in chapter 9 of Warrant and Proper Function.
II. Problems with Post-Classical Chisholmian Internalism
I propose to make four critical observations about this view. First, it is relatively uninformative; it tells us little about what warrant or positive epistemic status is. Second, this new view, while it holds on to the epistemic internalism of classical Chisholmian internalism, loses the principal philosophical motivation for internalism in moving away from deontology. Third, it is not the case (contrary to the post-classical Chisholm) that for a given belief B, there is a set S of evidence-bases such that necessarily, B has warrant for me if and only if it occurs in connection with a member of that set S. And consequently, fourth, it is a mistake to suppose that the warrant a belief enjoys for S can be understood as a function solely of the psychological properties S exemplifies.
A. Axiological Inspecificity
On the axiological view we don't really have an answer to the question what is warrant? We did on the classical view: there warrant was the relation in which a proposition stands to a person S when S can fulfill his epistemic duty by accepting it, the degree of warrant depending upon the degree to which S's accepting it is apt for the fulfillment of that duty. On the post-classical view, however, we are told only that warrant is or supervenes upon the occurrence of certain belief/evidence-base pairs—pairs ‹E, B›, such that a certain intrinsic value attaches to any state of affairs consisting in someone's having E and holding B.
But of course there are a thousand kinds of states of affairs that are intrinsically valuable: love, for example, but also happiness, kindness, peace, justice, truthfulness, faithfulness, loyalty, pleasure, courage, beauty, trustworthiness, moral dutifulness. There are also many epistemic intrinsic goods, such as believing the truth, speaking the truth, doing your epistemic duty and forming beliefs responsibly, holding coherent beliefs, holding beliefs for which you have sufficient evidence, believing what on sufficient reflection you would believe, being such that your beliefs are produced by reliable belief producing mechanisms, being such that your cognitive faculties are subject to no dysfunction, and many more. The post-classical view is thus much less specific than the classical one. On the classical view (as on the post-classical) warrant consists in an intrinsically valuable state of affairs, but the kind of intrinsic value is specified: it is that of the fulfillment of epistemic duty. On the post-classical view, a belief's having warrant is also a matter of its being an element in an intrinsically valuable state of affairs; but about the type of intrinsic value in question we are told only that it is not that of the fulfillment of duty. I look out the window and see a tiger lily; that is, I exemplify a certain set E of purely psychological properties (I am appeared to in a certain characteristic tiger-lily-like way and exemplify the rest of some set of purely psychological properties that go with seeing a tiger lily); I form the belief that I see a tiger lily. Now what is it that confers warrant upon this belief, under those conditions? Why does this belief have more warrant than, for example, the belief that I see a tiger? The axiological answer: the first situation has more intrinsic value than the second. But what kind of intrinsic value is it that is at issue? All we have by way of answer: it is not the the kind of intrinsic value exhibited by cases of duty fulfillment. But isn't this unduly uninformative? About all we seem to be able to say here is that the type of intrinsic value involved in warrant is the type of value displayed by a state of affairs consisting in someone's exemplifying a belief/evidence-base pair ‹E, B› which is such that when a person holds B while displaying E, then B has warrant for her.
Or perhaps the problem is not just that the answer is uninformative, but that it is unintuitive, implausible. It is certainly true that when I exemplify the properties in E, it is better that I believe that I see a tiger lily than a tiger. But why is it better? Well, of course the second belief is false and the first one true. More poignantly, if I form the second belief under those circumstances, I am at best in deep psychological trouble and at worst insane; forming that belief under those circumstances is a sign of a deeply disordered psyche. But these reasons for preferring the one situation to the other are not the relevant ones, according to the axiological view. What is relevant, according to that view, is simply that the state of affairs consisting in someone's exemplifying the set of properties in E and holding the first belief is intrinsically more valuable than someone's displaying those properties and holding the second. If there is such a difference in intrinsic value, it is not easy to detect.
B. Internalism Unmotivated
Second, I'd like to emphasize how very different this post-classical perspective is from classical Chisholmian internalism; and the heart and soul of this difference, as I see it, directly concerns the central notions of internalism. The axiological conception loses the connection with internalism and hence loses the philosophical motivation for it. As we saw in chapter 1, deontology implies internalism; but if we move away from deontology, if we suppose that what confers justification is not my being above reproach or acting responsibly, in accord with my duty, but rather a certain appropriateness of belief to evidence-base, then we lose that reason for accepting the internalist motifs. On the classical view, B has warrant for me to the degree that I can fulfill my epistemic duty—my duty to try to achieve epistemic excellence—by accepting it. This will be a duty, of course, only for “intellectual beings,” as Chisholm puts it—beings capable of grasping concepts and holding beliefs. Even then, however, it will be a (subjective) duty only for well-formed, nondefective intellectual beings. There may be those who by virtue of cognitive malfunction and through no fault of their own simply cannot see that it is their duty to undertake this project; such persons are not to be blamed for failing to undertake it. There may also be those who (perhaps like some French philosophers) believe it is their duty to try to spread as much darkness and confusion as possible; if such a person forms this belief nonculpably (by virtue of cognitive malfunction, perhaps, rather than a prideful effort pour épater le bourgeois) then she too is not to be blamed for failing to pursue the road of duty. But well-formed cognitively healthy persons will not, in the typical case, be able mistakenly but nonculpably to believe that epistemic duty requires a course of action incompatible with the pursuit of epistemic excellence.
Further, the ground of warrant, that which confers warrant upon a belief for me, is the aptness of my accepting that proposition for fulfilling my epistemic duty. Here again, if there is to be coincidence of objective and subjective duty, if we are guilty and worthy of blame for failing to do what is our objective duty, then (in the cases in question) well-formed human beings must not be able to believe falsely but nonculpably that a given candidate for belief is such that I can fulfill my epistemic duty by accepting it. But of course this is easily satisfied on the classical view. In the typical case the degree of that aptness will depend upon the degree to which the proposition in question seems to me to be true; and that is indeed something to which I ordinarily have special access.
These pressures toward internalism disappear when we turn to the post-classical Chisholm. Suppose we turn our backs on deontology; suppose we believe that what confers justification upon a belief B, for me, is not my having acted in accord with my duty (in forming and maintaining it) but rather some relation that holds between B and the set of purely psychological properties I exhibit. If duty is not involved, why suppose that to be justified I must have any sort of access at all, special or otherwise, either to the fact (if it is a fact) that this relation constitutes or confers warrant, or to the fact that a given belief bears that relation to the totality of my purely psychological properties? Suppose warrant is a relation between the totality of my purely psychological properties and my beliefs (a relation such that it is intrinsically good that it be exemplified); why think that I would have to have any sort of special access to this relation? This relation between evidence-base and belief might be such that we cannot ordinarily tell whether our beliefs and evidence-bases stand in it. It might be a relation such that it is easy enough for us to believe that it obtains when it does not. It might be a relation such that a person's evidence-base and beliefs can stand in it, even if that person can't so much as grasp or form a conception of it. It might be a kind of fittingness such that some kinds of creatures are able to detect its presence but others cannot, even though they are capable of belief. It might be a kind of fittingness that only an expert can detect; or it might be one such that we can detect it only by virtue of a certain stroke of epistemic luck—the sort of luck denied those unfortunates who are brains in a vat or deceived by Cartesian demons.
In “The Place of Epistemic Justification,” as we saw, Chisholm partially explains the locution ‘x is more reasonable for S at t than y’ by laying certain constraints upon it; one of these constraints is that “The relevant sense of reasonable belief is one which is such that a believer can ascertain by himself at any time which of his beliefs are reasonable for him at that time” (p. 86). But the question is: why say this? Why lay this down as an initial constraint on locating the notion to be explained? For the earlier classical view, there was a clear answer: the deontological connection. But for the later post-classical view this connection vanishes; and with it goes the reason for the epistemic internalist dimension of the view. The internalism lingers, like the smile of a Cheshire cat, after its raison d'être has disappeared.
C. No Set of Evidence-Bases Necessarily Connected with Warrant
According to Chisholm's later view, warrant, for one of my beliefs for a given belief B, consists in a certain relationship that holds between B and my purely psychological properties. More exactly, there are some evidence-bases such that if B occurs in conjunction with those evidence-bases, then it has warrant; but other evidence-bases such that if it occurs in connection with them, then it does not have warrant. As we have seen, this is so far compatible with a wide variety of views as to what sort of intrinsic value warrant involves, and hence with a wide variety of views as to what warrant is. It is compatible with coherentism, for example, according to which what counts for warrant is the appropriate relation among my beliefs, the rest of my purely psychological properties being irrelevant.13 It is also compatible with evidentialism, the view that what counts is the evidential relation between B and the relevant purely psychological properties. In Warrant and Proper Function, I argue that B has warrant for you only if it is produced by your cognitive faculties functioning properly, the degree of warrant it displays depending upon the degree to which you are inclined to accept it: the intrinsic value of which Chisholm speaks could also be the value attaching to the relation between B and your evidence-base when both are produced by properly functioning epistemic faculties.
But Chisholm takes a further step here, and a fateful one at that: he claims that for a given belief B, there is a set S of evidence-bases such that, necessarily, B has warrant for me if and only if it occurs in connection with a member of that set S. What is distinctive is the idea that for any belief B there is a certain set of evidence-bases S such that B's occurring in connection with a member of S is both necessary and sufficient, in the broadly logical sense, for B's having warrant. Take a given belief B you now hold: the one thing relevant to the question whether B has warrant for you is the question what purely psychological properties you now display; there will be a set of purely psychological properties such that having some member of that set as your evidence-base will be both necessary and sufficient (in the broadly logical sense) for B's having warrant for you.
I think this is mistaken. Before trying to explain why, however, we must state the view a bit more exactly. Begin by recalling Chisholm's reduction of epistemic concepts to the concept of intrinsic value:
(ED8) Believing p is epistemically preferable for S to believing q = def. Those of S's purely psychological properties which do not include believing p and believing q are necessarily such that having those properties and believing p is intrinsically preferable to having those properties and believing q.
Now consider what is involved in (ED8). Say that a maximal set S of purely psychological properties is a set of psychological properties such that (1) it is possible for a cognizer to have every property in S, and (2) it is not possible for a cognizer to have every property in any superset of S; and say that a maximal psychological property P is the property of having every property in some maximal set of purely psychological properties. Further (where B is a belief) say that a maximal set diminished with respect to B is the result of deleting from a maximal set every property that includes believing B; and say that a maximal psychological property diminished with respect to B is the property, for some maximal set of psychological properties diminished with respect to B, of having every property in that set. Of course there will be pairs ‹B, P› whose second member is a maximal psychological property diminished with respect to B and whose first member is B; call these epistemic pairs. According to Chisholm, some epistemic pairs display more by way of intrinsic value than others. There will be true (indeed, necessarily true) propositions of the form ‹Bi, Px› has more intrinsic value than ‹Bj, Py›; and if you exemplify ‹Bi, Px› and I exemplify ‹Bj, Py›, then Bi will have more warrant for me than Bj does for you. There will be an ordering (perhaps only partial) of epistemic pairs in terms of intrinsic value; this ordering will induce another ordering in terms of level or degree of intrinsic value (the details of the ordering can be adjusted to suit). So for each epistemic pair, there will be a degree or level of intrinsic value; and if I exemplify a given pair ‹Bx, Pi›, then Bx will have for me the degree of warrant appropriately related to the degree of intrinsic value displayed by ‹Bx, Pi›. More precisely (if more pedantically) for any belief B and degree of warrant d, there is a set C of epistemic pairs (each of whose first members is B) such that, necessarily, for any person S and time t, B has d for S at t if and only if S then exemplifies some member of C. Alternatively, for any belief B and degree of warrant d, there is a set C∗ of evidence-bases such that, necessarily, for any person S and time t, B has d for S at t if and only if some member of C∗ is S's evidence-base at t.14 Less precisely but more intelligibly, the degree of warrant now enjoyed by one of my beliefs depends only upon which purely psychological properties I now exemplify.
This is the claim I mean to dispute; I shall argue that, for a given belief of mine and a given degree of warrant, there is no set of evidence-bases such that, necessarily, the belief displays that degree of warrant if and only if my evidence-base is a member of that set. I don't mean to dispute the claim that some belief/evidence-base pairs display more intrinsic value than other such pairs; that is as may be. But even if this claim is true, that value is not what constitutes warrant. I shall argue that many of the epistemic pairs ‹B, P› I exemplify are such that (1) B has a great deal of warrant (perhaps a degree sufficient for knowledge), and (2) there are other conditions in which I could exemplify the same pair but B have little warrant for me. I shall also argue that there are many epistemic pairs ‹B, P› I exemplify such that (1) B has little or no warrant for me, and (2) there are other conditions in which I could exemplify the same pair but B have a great deal of warrant for me. But if I am right in these arguments, then it is false that for any belief B there is a set of evidence-bases such that, necessarily, B has warrant if and only if it is accompanied by one of those evidence-bases. It is false that all that counts, in determining the degree of warrant a belief of mine has, is the evidence-base I then display.
So on to the arguments. First, note that our cognitive equipment could have been quite different from what it is. Clearly, we (or, to beg no questions, a species similar to us) could have been so constructed that quite different phenomenal properties would have been associated with a given belief when it constitutes an item of high epistemic status for us. We could have had a sensory apparatus of a very different sort: for example, we could have been endowed with the sort of sense by which a pit viper can determine the distance and direction of a mouse by detecting its body heat; we could have had the kind of directional awareness that enables birds to navigate vast distances without so much as a glance at a map; we could have had an awareness of the force and direction of the earth's magnetic field that according to some is bestowed upon sharks; we could have had the sort of acoustically based sensory equipment hat enables a bat to thread its way with great precision and at high speed through a nasty set of obstacles in a completely dark cave. And of course quite different phenomenology could go with such different senses, including psychological properties no human being has ever exemplified. So the set of pairs ‹B, P› in the hierarchy will include pairs whose first members include phenomenal properties no human being has ever enjoyed. Similarly, there may be or could be cognizers who grasp a much broader range of propositions than we do so that they grasp and believe and reject propositions completely beyond our ken; hence the pairs in the hierarchy will also include some whose second member is a proposition no human being has ever grasped. Of course this is no problem, so far, for Chisholm; it is only that the field of epistemic pairs, for a given degree of warrant, is broader than we might have initially thought.
Next, note that we learn what beliefs to form under given phenomenal conditions, what beliefs to form for a given (complex) way of being appeared to. We learn how to judge distance, for example; and we learn that, despite appearances, railroad tracks do not really converge in the distance and that people and automobiles are not actually smaller when viewed from the top of a tall tower. What we learn depends upon circumstances and upon contingent circumstances; our circumstances could have been such that we should have learned something quite different. A large oak tree viewed at fifty yards under standard conditions displays a certain phenomenal appearance, as does a mountain goat on a dark crag viewed at 350 yards under good lighting conditions. Furthermore, if I am practiced at viewing mountain goats, I can take a good look at one 350 yards or so away and thereby know both that it is a mountain goat and that indeed it is about 350 yards away. This takes a certain amount of practice or learning; a mere tyro might guess that what he sees is a goat at 350 yards, but the seasoned old mountain woman knows. It isn't necessary, however, that the phenomenal properties that in fact go with seeing a mountain goat at 350 yards should confer warrant on the belief that it is 350 yards distant. If different atmospheric or lighting conditions were the norm, or if our sensory apparatus had been differently constructed, that way of appearing—the way that in fact goes with seeing a mountain goat at 350 yards—might well have gone with the belief that it was 100 yards off. What confers positive epistemic status upon the belief in question under the phenomenal circumstances in question is not just those phenomenal circumstances; what you have learned is also relevant.
But then for Chisholm's view to be correct, what you have thus learned will have to be represented or included, somehow, in the purely psychological properties you display at a given time. Perhaps you will have, at any such time, a very large set of beliefs of the sort: when it looks like that, it is a mountain goat at 250 yards; when it looks like that∗, it is a mountain goat at 300 yards, when it looks like that∗∗, it is a mountain goat at 500 yards, and so on; or perhaps what you have learned will be represented by a sort of feeling of appropriateness, or fit, between its looking like that∗ and the judgment that it is a mountain goat 300 yards away, or perhaps in still other ways. Assume, for the sake of simplicity, that what you have learned is represented by beliefs of this sort. Now from a post-classical Chisholmian perspective, what is it about your evidence-base that confers warrant upon your belief that you see a mountain goat at 300 yards? Such facts, of course, as that it looks like that∗, and that you believe that when it looks like that∗ it is a mountain goat at 300 yards. But what about that latter belief, the belief that when it looks like that∗ it is a mountain goat at 300 yards? What is it that confers warrant upon it? It is hard to see how it could be some relation between that belief and the rest of the purely psychological properties I display at the time in question. Even if a mere tyro had the same belief, it would not have the same degree of warrant. It looks as if what counts for warrant, on this axiological view, must be more than a relation between the belief in question and the purely psychological properties you exhibit at the time in question; the course of your past experience is also relevant.
But now to the heart of the matter. I propose to argue that the degree of warrant one of your beliefs B has does not depend merely upon the purely psychological properties you display; in different circumstances, the same epistemic pair ‹B, P› can be associated with vastly different degrees of warrant. Consider first so-called a priori beliefs—such beliefs as 2 + 1 = 3, 31 − 6 = 25, Complex Constructive Dilemma is a valid argument form, sets are neither true nor false, and if all men are mortal and Socrates is a man, then Socrates is mortal. Each of these is a proposition I believe, and each, I believe, has a great deal of warrant for me. Now according to post-classical Chisholmianism, what confers warrant upon these beliefs for me is their being associated with the right evidence-bases. But what could it be, about my purely psychological properties, that thus confers warrant upon them? The Chisholmian answer, I suppose, is that it is the phenomenal properties to which we refer when we say that we can simply see that the proposition in question is true (and perhaps what we simply see is not just that it is true, but that it must be true). Perhaps (and then again perhaps not) this is a matter of the clarity and distinctness of which Descartes spoke. Locke spoke in the same connection of an “evident luster”; a self-evident proposition, he says, “displays a kind of clarity and brightness to the attentive mind.” Perhaps Descartes and Locke don't have the matter precisely right; perhaps the phenomenology involved is not exactly clarity or brightness or luster.15 Still, something like what they say must be true from the post-classical perspective. If what confers warrant on a simple self-evident belief B for me is the intrinsic value of its conjunction with certain other purely psychological properties, then presumably those other purely psychological properties will be the phenomenal properties I exemplify when I contemplate or attentively consider B.
But of course we, or creatures similar to us, could have been constructed quite differently. Consider, for example, the first 25 primes. Most of us cannot simply see, with respect to each of them, that it is indeed prime; we need to calculate, some of us in fact requiring pencil and paper. But of course we could have been constructed in such a way that we could simply see, with respect to any member of this set, that it is prime; perhaps in fact there are angels or Alpha Centaurians who can do exactly that. Such creatures would have the same sort of knowledge of these propositions as most of us have of the simplest propositions of elementary arithmetic. So consider an epistemic pair ‹B, P› where B is the belief, say, that 67 is a prime, where P is a purely psychological property reduced with respect to B, and where ‹B, P› is such that I could know B if P were my evidence-base; there are accordingly possible circumstances in which I exemplify ‹B, P› and know B—circumstances in which B has a very high degree of warrant for me. I want to argue that there are other possible circumstances in which I exemplify ‹B, P› and am such that B has very little warrant for me—but if so, then, clearly enough, for B there is no set of evidence-bases such that necessarily, B has warrant for me if and only if it is accompanied by some member of that set.
So suppose I am captured by a group of unscrupulous but extremely knowledgeable Alpha Centaurian superscientists intent upon a cognitive experiment (or suppose that I am a victim of a Cartesian demon who delights in deception). In the course of conducting their experiment, the scientists (or demon) modify my cognitive faculties. Consider the propositions of the form n is prime where n is a natural number between 23 and 200 (inclusive); my captors bring it about that I believe every third proposition of this form and disbelieve the rest. Thus I believe that neither 23 nor 24 is prime, but do believe that 25 is. Further, for each of those propositions I believe—25 is prime, 28 is prime,… 67 is prime,… 199 is prime—they so modify my faculties that when I consider it, I undergo the very sort of phenomenology that in fact goes with simply seeing that, say, 5 is prime. In the case of each of these propositions it seems to me that I can simply see that it is true in just the way I can simply see that 2 + 1 = 3 is true. Similarly for the ones I believe to be false; each of them is such that when I consider it I undergo the phenomenology that goes with seeing, for example, that 31 + 32 = 277 is false. I therefore think I can simply see that 67 is prime, just as I can simply see that 2 + 1 = 3 and just as I think I can simply see that 5 is prime. When I consider 67 is prime, it has for me precisely the phenomenology that goes with simply seeing that a simple arithmetical proposition is true. But of course my faculties have been modified so that they no longer function properly; in particular, most of the propositions of the form n is prime that I think I can simply see to be true are false, and most of the ones I think I can simply see to be false are true. It is, so to speak, only by accident that I have the right phenomenology with respect to 67 is prime; my captors could just as well have given me the phenomenology that goes with seeing that a proposition of the form n is prime is false.
It is clear, I think, that under these conditions 67 is prime has very little by way of warrant or positive epistemic status for me. Although it is true and necessarily true, I surely do not know that it is. So if I were in this condition of epistemic malfunction—brought about by a malicious Cartesian demon, inquisitive Alpha Centaurians, or perhaps a rare brain lesion—then I could display the phenomenology that goes with just seeing that 67 is prime, but nonetheless be such that that proposition has little or no positive epistemic status for me. But then consider that epistemic pair ‹B, P› where B is the belief that 67 is prime and P is a purely psychological property reduced with respect to B and where the pair is such that I could know B if P were my evidence-base. Clearly I could exemplify that pair but be such that B has very little by way of warrant for me. But of course I could also exemplify that pair and be such that B has a great deal of warrant for me. So the degree of warrant a belief has for me does not depend merely upon the purely psychological properties I exemplify at the time in question. It also depends upon whether or not my faculties are functioning properly in producing that belief in me then. And if that is so, there is no set of evidence-bases such that, necessarily, the belief in question has warrant if and only if it is associated with a member of the set.
Let me be clear about the logic of the situation. According to the post-classical Chisholm, the only thing that can confer warrant or positive epistemic status upon a belief B for me is my exemplifying an appropriate evidence-base. For any belief B and each degree of warrant d, there is a set E of evidence-bases—that is, maximal purely psychological properties reduced with respect to B—such that necessarily, for any person S, B has d for S if and only if S exemplifies a member of E. What we have seen so far is that the evidence-bases that typically accompany high warrant for us are not sufficient for such warrant; in other conditions (pathological conditions) a person might exemplify the evidence-base but have very little by way of warrant for B. It is worth noting (though not inconsistent with what Chisholm says)16 that we can also proceed in the opposite direction: having one of the evidence-bases that for us accompany high warrant for a belief B is also not necessary for B's having high warrant. We can see this as follows. For us, the phenomenology that accompanies seeing, for example, that 7 + 5 = 12, is (on post-classical Chisholmianism) a matter of its seeming obvious or clearly true not only that the proposition in question is true but that it must be true, couldn't be otherwise. But presumably this is not necessary; there could be creatures who see that 7 + 5 = 12 is true but do not see that it couldn't be false; perhaps, indeed, they lack the conception of necessary truth. Surely God could make creatures whose phenomenology, with respect to truths of this sort, differed from ours in just this way; and perhaps he could compensate them, for their failure to perceive the necessity of these propositions, by conferring upon them the ability to see directly (without inference or calculation) the truth of many more arithmetical propositions than we can.
What we have here is a whole class of beliefs such that the degree of warrant they have for S does not depend merely upon S's exemplifying the right purely psychological properties; we must also ask whether S's faculties are functioning properly. But of course the same holds with respect to beliefs of other sorts. Consider memory beliefs: on Chisholm's view what confers warrant upon a given memory belief—that this morning I had an orange for breakfast, for example—is the sort of phenomenology I undergo when I consider the question what I had for breakfast. When I ask myself this question, that belief suddenly appears, simply occurs to me, pops into my mind, we might say, along with a sort of past-tinged phenomenology that is hard to describe but familiar to all. But of course a Cartesian demon or Alpha Centaurian cognitive scientist could modify my memory faculties. Say that my memory field is the set of propositions that constitute what I at present seem to remember, together with the negations of those propositions.
Now imagine that a Cartesian demon, acting out of sheer malicious whimsy alters my faculties by redistributing over my memory field the phenomenology that in fact goes with propositions I do indeed remember. The distribution is completely random and is such that no more than a quarter of the propositions I seem to remember are true. Then what I seem to remember, even if true, will have very little by way of warrant or positive epistemic status for me, although of course I will be deontologically justified in accepting it. Such beliefs will certainly not constitute knowledge. The reason is that even if, as it happens, it is a true belief that is accompanied by the phenomenology in question, it is just by epistemic accident that this phenomenology goes with that belief; the demon could just as well have conferred that phenomenology upon a false belief. So these beliefs will have very little warrant for me—even though under other circumstances, when I exemplify the same epistemic pairs, the true beliefs in question do constitute knowledge. What we see once more is that exemplifying one of the evidence-bases that in fact accompany a belief B's being of high warrant for me is not sufficient for that high degree of warrant; more is required, and the more has to do with the proper function of cognitive equipment.
The same point can be made with respect to perceptual beliefs. Consider the evidence-bases that in fact accompany perceptual beliefs for which I have a high degree of warrant: I think we can see that having one of those evidence-bases is not sufficient, in the broadly logical sense, for those beliefs having that degree of warrant. Suppose, once more, I am being arbitrarily manipulated by an Alpha Centaurian or Cartesian demon or am subject to some other pathology-inducing condition, and consider the phenomenology, the purely psychological properties, that in fact go with perceiving a large oak tree at 40 yards: suppose the Alpha Centaurians give me those properties at random intervals, with no correlation at all with my being in the presence of oak trees. Naturally enough, on those occasions I believe that I perceive an oak. These beliefs, however, will surely have little by way of warrant for me, despite the fact that they are accompanied by evidence-bases that in other, more happy circumstances accompany that belief's having a high degree of warrant for me. In this case as in the others I will of course be deontologically justified in accepting the beliefs I do accept, and I will also undergo the right phenomenology; but due to the pathology those beliefs will have very little by way of warrant for me—much less than such beliefs have for someone whose faculties are functioning properly.
Now in the conditions envisaged, my belief that I perceive an oak is false; on those occasions I do not perceive a tree at all. But we can easily modify the example to get a similar one where the belief in question is true. Suppose I see a quick furry blur bound across the space between a couple of trees in my back yard: I may form the belief a squirrel just ran from that tree to that tree. Clearly, once more, the Cartesian demon could arrange for me to have that sort of experience on a completely random basis; if it should happen to occur just after a squirrel did run from that tree to that tree, then, though the belief is true, it has next to no warrant. Surely it does not constitute knowledge, as it might if my faculties were functioning properly.
I have been arguing that, pace the post-classical Chisholm, it is false that for any belief B I might have, there is a set of evidence-bases E such that, necessarily, B has warrant for me if and only if it is a member of E. I argued this by finding beliefs B for which there are no such Es; the argument proceeded by way of finding a B and an evidence-base E such that in normal conditions one who holds B and has E as evidence-base is such that B has warrant for her, but in other conditions, conditions of cognitive malfunction, a person might hold B, have E as her evidence-base, and be such that B has no warrant for her. Another way to argue the same point is to find a belief B and evidence-base E such that, (1) as we are in fact constituted, one who holds B and has E is such that B has no warrant for her, but also (2) in other conditions a person who held B and had E could be such that B did have warrant for her.
So consider the unfortunate Paul of the last chapter (p. 42), who formed beliefs of the wrong modality: when appeared to in the church-bell fashion, he formed the belief that something was appearing to him in that fashion, and that it was orange. Now as things presently stand, it is clear, I think, that beliefs of these sorts do not have warrant. Paul's pathologically induced belief has no warrant, and even if it happened, on a given occasion, to be true, it would not constitute anything like knowledge. But of course things could have gone differently. Suppose God (or evolution, or both) had designed human beings in quite a different fashion—or, to avoid avoidable questions about what is involved in being a human being, suppose God had created a species capable of knowledge whose members were a lot like human beings but differed in certain crucial respects. Just as we are by nature such that when appeared to by something that is red, we form the belief that we are appeared to in that way by something that is red, so these creatures are by nature such that when appeared to in the church-bell fashion, they form the belief that they are appeared to that way by something that is orange. Imagine further that although these beings are often appeared to in that orange fashion, they inhabit a planet on which they seldom (if ever) visually perceive that an object is orange; atmospheric conditions make that for the most part impossible. (It is only once a year or so that one actually sees something that is orange and sees that it is orange.) Add that as a matter of fact nearly everything on this planet that makes the church-bell sound in question is orange. Now imagine that there is a certain common but rarely visible orange bird that makes the church-bell sound. When the inhabitants of this planet catch a glimpse of this bird (without seeing its color) and hear it make that sound, they form the belief that there is something appearing to them in the church-bell fashion and that it is orange. Why wouldn't this belief have a high degree of warrant for them—indeed, why couldn't it constitute knowledge? And this even though in Paul's case his similar belief and evidence-base are such that the latter confers no warrant upon the former?
D. Psychological Properties Insufficient for Warrant
I argued in the preceding section that the post-classical Chisholm is mistaken in arguing that for any belief B I might have, there is a set E of evidence-bases such that, necessarily, B has warrant for me if and only if it is a member of E. But of course this shows that warrant is not a function simply of what epistemic pairs I exemplify. Given that I hold a belief B, it is not the case that whether B has warrant for me depends solely on the purely psychological properties I display; purely psychological properties are not the only thing relevant. But then we should expect that Chisholm's post-classical account of knowledge will be mistaken. And I think it is easy to see that it is.
According to the post-classical Chisholm, the basic component of the concept of warrant is evidence. Take a case where you know something B—that 7 + 5 = 12, for example, or that you had an orange for breakfast. In any such case you will have an appropriate evidence-base E, which will make B evident for you; the evidence itself is really a degree of intrinsic value, the degree that the combination of such evidence-bases as E with such beliefs as B displays. Evidence is the basic component of warrant, for the post-classical Chisholm, but it isn't the whole shooting match; Gettier situations may bring it about that a belief is evident for you but does not constitute knowledge for you. Chisholm proposes a codicil to deal with Gettier problems. When a belief is evident for you, we may say that your evidence-base makes it evident for you. Now an evidence-base can sometimes make a false proposition evident for you (thus Smith might give me overwhelming inductive evidence that he owns a Ford when the fact is he's just amusing himself at my expense and owns no car at all). But then, says Chisholm, your belief B constitutes knowledge for you if and only if your evidence-base makes B evident for you, and furthermore your evidence-base makes no false proposition evident for you.17
It is easy to see, I think, that this account is insufficient; we need only reflect on the examples I have given. Take the example (p. 59) where, for every third proposition of the form n is prime (n between 23 and 200), my Alpha Centaurian captors playfully give me the phenomenology that goes with simply seeing that a number is prime. I consider the proposition 67 is prime; on the post-classical Chisholm's account of knowledge, this proposition is then evident for me (since it is accompanied by the right purely psychological properties). It is also true, and my evidence-base need make no false proposition evident for me; but then on the above account it constitutes knowledge for me. But surely it doesn't. Or take the case (p. 62) where the demon every now and then and at random gives me the phenomenology that normally goes with the belief a squirrel just ran from that tree to that tree. If he gives me that phenomenology just after a squirrel did in fact run from that tree to that tree, the belief will be true and evident, and the evidence-base need make no false proposition evident. But then, once more, that belief constitutes knowledge; and surely it does not.
Justification most properly so called, as we saw in chapter 1, is a matter of deontology; it crucially involves duty, permission, obligation, and the like. As we saw in chapter 2, justification so construed is wholly insufficient for warrant, and also unnecessary for it. What we have seen in this chapter is that justification taken the post-classical Chisholmian way—as a sort of intrinsic value that necessarily attaches to certain belief/evidence-base pairs—is also nowhere nearly sufficient for warrant. So we have two quite different Chisholmian accounts of justification, and neither is anywhere nearly sufficient for warrant. Justification construed either of these two ways is a specification of that vaguer and more capacious notion of justification I called ‘broad justification’ according to which a proposition is justified for you if you believe it and everything ‘downstream of experience’ is ‘going properly’. What we have so far seen, then, is that neither of these two Chisholmian versions of broad justification is nearly sufficient for warrant.
We shall return to broad justification and to other versions of it; but we can already see in a preliminary way that no such version is at all likely to be correct. That is because they all neglect the important contribution of the epistemic environment to warrant. With deontological justification (justification most properly so called), all that counts is whether I do my epistemic duty; the vagaries of my cognitive environment are not relevant. On Chisholm's post-classical conception, all that counts for the justification of a belief is the right set of purely psychological properties; again, my cognitive environment is irrelevant. But neither sort, as we have seen, is sufficient for warrant; and the insufficiency has to do with the neglect of the cognitive environment. Suppose I am transported instantly and without my knowledge to a distant planet where the cognitive environment is very different (elephants, for example, are invisible, but give off a sort of radiation that causes human beings to be appeared to as if a trumpet is sounding nearby), so that while I do acquire a few true beliefs, it is just by happy cognitive accident. Suppose further that everything downstream of experience is going properly; do I really have knowledge of the few true beliefs I accidentally acquire? We shall have to look into this matter further; but initially one thinks not.
Our present concern, however, has been with the version of broad justificationism developed by the post-classical Chisholm. What we have seen is that what determines the warrant a belief has for me on a given occasion is not simply the evidence-base, the purely psychological properties I exemplify then. Considerations of proper function and pathology are also relevant. The later Chisholm's views as to what constitutes warrant or positive epistemic status, like those of the earlier Chisholm, are not the true story. The epistemic principles he proposes, particularly if accompanied by a codicil to the effect that S's cognitive faculties are working properly, may be correct or nearly so; but the account of the nature of warrant is fundamentally flawed. In Warrant and proper Function I shall try to develop a better answer.
Chisholm's varieties of internalism, therefore, are not adequate. But there are other varieties of internalism. In particular, there is coherentism. Coherentism, as I said earlier, is perhaps best thought of as a sort of special case of post-classical Chisholmianism; strictly speaking, therefore, we already have the materials for seeing that it is inadequate. That is no substitute, however, for detailed consideration of it, and it is to this that we turn in the next chapter.