Find EnlightenmentContentsPrevious Chapter



I. Meta-Skepticism.

Variations on the Meta-Regress Argument can, like the ordinary regress argument, be used to support various conclusions, depending on how (or whether) one proposes to end the regress. The ordinary regress argument's typical conclusion is foundationalism. An alternative conclusion, which is usually considered an option particularly to be avoided, is skepticism, e.g., the view that no belief is justified. Now it appears that, after an extended presentation of my own Meta-Regress Argument, I have arrived not at meta-foundationalism but at a meta-skeptical conclusion. But this is not quite obvious and will require further investigation.

So the purpose of this chapter is to determine what conclusions ought to be drawn from the results of the first three chapters. Among the recommended conclusions to be advanced is that meta-skepticism can be avoided. Toward the end of avoiding it, I will advance a theory of rationality according to which it can be plausibly claimed that certain fundamental standards might be rationally, if not justifiably, believed, and that those standards license arguments that support (and justify) nonbasic standards. A number of highly contentious issues and objections will have to be addressed along the way.

Let us begin by formulating what may be advanced, uncontroversially, as one significant conclusion of Chapters 1-3. (It would be controversial to draw certain other conclusions, which I shall nonetheless draw later on in this chapter; but we will start with the following less-controversially-drawn conclusion.)

It might well be that there are many j-standards that receive support from arguments the premises of which are themselves justified in accordance with other j-standards. But if we press the issue, we encounter a regress. And as I have argued, we cannot accept arguments alleging that standards support themselves, either via a single argument or many. Moreover, as advanced in the Chapter 3, to suppose that it is precisely the premises of an argument for a basic j-standard that are justified but somehow not in accordance with a standard or justifiers or to suppose the same of the standard itself is ad hoc and probably absurd as well. So when we ask, of a given standard, "What is the justification of this standard?" we are stuck saying, "Ultimately, nothing. There might be some immediate supporting arguments, but if the issue is pressed there is nothing that ultimately justifies this or any j-standard."

This claim can be stated more precisely as follows:

(C) For any given justification standard, J, acceptance of it is either not justified, or ultimately receives support only from beliefs that are themselves not justified (regardless of how many intervening beliefs there might be between the ultimately supporting beliefs and the standard).

This is what I take Chs. 1-3 to have established.

I think most would be willing, without further ado, to call (C) an unacceptably skeptical conclusion; but that charge is, I think, hasty. Let me explain. Skepticism comes in a variety of forms, two of which are relevant to formulating the charge:[146]

(SK) Global skepticism: no belief is epistemically justified.

(MSK) Meta-skepticism: no belief in a justification standard is epistemically justified.

A particularly strong indictment of (C) would involve claiming that, first, it entails (MSK), and second, since (MSK) entails (SK), (C) also entails (SK), i.e., global skepticism. Let us see what merit there is to these charges, beginning with the second, stronger charge.

While (SK) obviously entails (MSK), the entailment in the opposite direction is less obvious. In his famous diallelus argument, Sextus Empiricus appears to have assumed this opposite entailment;[147] that is, he assumed that in order to have knowledge, one must have a criterion according to which that which one claims to know is true, since the use of such a criterion would be the only way to demonstrate knowledge. And if one cannot demonstrate knowledge, Sextus assumed, one cannot legitimately claim to have it. Hence, failure to defend any criterion of truth would indicate that one does not know anything (or at least that one cannot legitimately claim to know anything).

But Sextus' view appears to assume that, in order to be justified in holding a belief, one must be aware of that in virtue of which one is justified in holding the belief. But as self-supportists are keen to point out, and as Alston persuasively argues in "Level Confusions in Epistemology,"[148] there is little support for this assumption.

Even if (C) is innocent of entailing global skepticism, meta-skepticism would be burden enough, at least as far as philosophers are concerned: it would remove a good reason to pursue a major part of epistemology. If we believed that no j-standards could be justified, what would be the point of debating about different theories of justification? There would be no point that was related to truth, at least. We might still have political motivations or other motivations not related to truth-seeking. But I believe most (not all) philosophers today are not nearly as interested in the mere political impact of their philosophical views as they are in whether their views are true. Accordingly they have reason to try to avoid meta-skepticism.

David Hume and a number of philosophers following after him have made the point (in various ways) that skepticism, as a philosophical theory, need not have any noticeable effect on our everyday attitudes, views, and decisions; hence they reject one objection to skepticism, viz., that it would render ordinary life impossible. So one might suggest that a similar argument is available in defense of meta-skepticism.

Whatever the merits of this sort of argument, the meta-skeptic cannot avail himself of it. Meta-skepticism is a meta-level theory, about philosophical claims (j-standards); to accept this theory seriously is to believe, sincerely, that no j-standard (among which are included some of the most cherished theories of epistemology) is justified. If no standard is justified, then the activity of offering an epistemic justification for a standard will always fail to demonstrate the standard's justification. Just that activity is what exercises philosophers engaged in the theory of justification; hence, their activity is doomed to failure, if meta-skepticism is correct.

Traditional epistemologists will accept, then, that meta-skepticism must be avoided. Essentially, if we are duly persuaded of (C), then to avoid meta-skepticism, we must find a way to hold onto the following conjunctive claim:

(1) All beliefs in justification standards are either not justified, or ultimately supported by beliefs that are themselves not justified; and some beliefs in justification standards are epistemically justified.

Evidently, for the conjuncts making up (1) to be consistent, it must be true that some beliefs in j-standards are epistemically justified but are ultimately supported (only) by beliefs that are themselves not justified. In other words, some justified beliefs receive their justification from beliefs that are not justified. That is contrary to what might appear to be one of the most basic canons of epistemology, to wit:

(2) If S is justified in the (nonbasic) belief that p, then S has some other justified belief (or beliefs) q that supports the belief that p.

Indeed, this might be regarded as following from a few uncontroversial claims, including an account of `nonbasic belief', such as the following:

Def. The belief that p is nonbasic for S iff p is justified only by other of S's beliefs if p is justified for S at all.

Consider also in this context the supports constraint, advanced in Chapter 2:

The alleged justifiers for a belief (whatever sort of thing they might be) support the belief they justify.

From these two claims it follows that nonbasic beliefs are supported by other beliefs, which is not precisely what (2) says. What (2) adds is that the justifiers that support a nonbasic belief are themselves justified beliefs.

It seems intuitive that nonbasic beliefs possess whatever degree of justification they have due in at least large part to the degree of justification their supporting beliefs have. Moreover, a belief that is not justified at all, it appears, cannot "give" any justification to other beliefs. If the belief that q is not justified, it cannot offer any justificatory support to the belief that p. Hence, if all j-standards are either not justified, or ultimately supported by beliefs that are themselves not justified, then no belief in j-standards is epistemically justified. The central conclusion of Chapters 1-3, (C), would appear to entail meta-skepticism.

It will take a considerable effort to show that appearances in this case, however plausible, are deceptive. Indeed, I wish the present chapter only to be construed as an attempt to show a possible way around meta-skepticism, not as a definitive refutation.

Here is the first step in the effort: it could be a category mistake to apply the concept of justification to some beliefs.

`Category mistake' describes the error of ascribing an attribute to an item in a category to which the attribute does not (literally) apply at all; for example, "My decision weighs five pounds" or "The sun is just." The meta-skeptical argument just stated fails to acknowledge the possibility that the concept of justification might not apply to a certain class of belief at all.[149]

An example of a similar sort of category mistake, arguably, would be to insist that the law of bivalence applies to all ordinary language statements. For example, if I were to insist that the claim "She is taller" is true (or false), when it is unclear who "she" denotes and who or what she is taller than, I could very plausibly be said to be committing a category error in the sense that I am trying to attribute truth to a member of a category to which the attribute does not properly apply. To the contrary, one may argue, ordinary vague (or otherwise unclear) statements are neither true nor false; the concept of truth does not apply to claims that are sufficiently unclear.

So one might identify three general categories of belief: the justified, the unjustified, and those such that it is a mistake to call them either justified or unjustified (nonjustified beliefs).[150]

Just suppose, then, that acceptance of a standard about (for example) the justification of some particular sensory beliefs is claimed to be nonjustified (neither justified nor unjustified) – that the concept of justification does not properly apply to that belief. Then one could well say that the belief in the standard is not justified; but one could not (accurately) say that the belief is unjustified. The question at issue is whether such a belief could possibly give justificatory support to other beliefs.

The meta-skeptic might insist that, regardless of whether the concept does or does not apply, (2) holds: a nonbasic belief is not justified unless it is supported by a justified belief. Admittedly, that insistence has significant force.

But give due consideration to this possibility: while it is on all hands unacceptable to say that an unjustified belief (i.e., one that has a certain kind of negative epistemic status) gives justificatory support to a belief, it might be possible that a belief, of which it makes no sense to say it is justified or not, gives the required epistemic support to another belief by having a different kind of positive epistemic status. That is, there might be some belief that is nonjustified, neither justified nor unjustified, because the concept "justification" does not apply to it; nonetheless, it is rationally believed, or warranted, etc. And suppose that it is the sort of belief that can, in the terminology of Chapter 1, properly "license" arguments for various other beliefs, including beliefs in j-standards. In that case, (C) would not be contradicted: a derivative j-standard "ultimately receives support from beliefs that are themselves not justified," and the world is safe for the theory of justification.

This might seem to be a silly suggestion, contrived to avoid skepticism. The challenge before us is to make this suggestion plausible, particularly in the face of the admittedly persuasive insistence that a nonbasic belief is not justified unless it is supported by a justified belief. As it turns out, closely related (or supporting) suggestions were made by some of the most revered names in epistemology, including Reid, Moore, Wittgenstein, and Strawson, and so their names will appear, as appropriate, in what follows. I will develop a meta-epistemological theory most heavily indebted to the work of the great Scots philosopher Thomas Reid. This work will help explain how, for all the problems raised in Chapters 1-3, one might still be in a position to reject meta-skepticism.

II. Toward a Theory of Minimal Epistemic Rationality.

How might it be that nonjustified beliefs can support justified beliefs? The following seems uncontrovertible: if a belief lacks any positive epistemic status, it cannot give epistemic support to another belief. In other words:

(3) If S is justified in the (nonbasic) belief that p, then S has some other belief (or body of beliefs) q, that both possesses some positive epistemic status and supports the belief that p.

Contrast this with (2) from Section I above, which states that all justified nonbasic beliefs are supported by justified beliefs. To the contrary, it can be suggested that some nonbasic, justified beliefs are adequately supported by beliefs that are nonjustified but rational.[151] The rationality of a nonjustified belief might be adequate to give epistemic support to an entire superstructure of justified beliefs.

A number of questions about this proposal are apt to leap to mind, but one should be dealt with right away, at least briefly. Namely, there is an epistemic circularity problem about justification; why not think that there would be an equally devastating problem for rationality?

The reason is that – on some notions of rational belief, and there are many – some rational beliefs, in order to be rational, need not be supported at all. As the Supports Argument from Chapter 2 indicates, the fact that justification requires noncircular support makes epistemic circularity unavoidably vicious. But some kind of rationality (or warrant, etc.) could be a variety of positive epistemic status that does not require support at all, noncircular or otherwise.

As an example (to be developed anon), a belief might be rational, in a relevant sense of `rational', due precisely to the fact that it is a spontaneous result of a properly functioning doxastic practice, such as the practice of looking closely. The belief is constituted as rational on account of its source.[152] In that case, its origin does not epistemically support its claim to rationality in the way that a basic belief allegedly receives justificatory support from a perceptual state. Properly speaking, one might argue, the belief is not epistemically supported at all and remains nonjustified. The fact that it results from a properly functioning practice merely accounts for its rationality.

But evidently, the notion of rationality at work here requires some explication. So our task is to develop an (at least minimally) adequate theory of epistemic rationality, on which a nonjustified belief (particularly, in a standard that might license an argument for a j-standard) might be rational. The central constraint on this theory is that it should permit us to argue that some rational but nonjustified beliefs offer adequate epistemic support to justified, nonbasic beliefs.

It is worth considering whether the one recent prominent theory of epistemic rationality, due to Richard Foley, might serve the purpose.[153] His theory in his own words is:

it is epistemically rational for a person to believe a proposition just if there is a way of arguing for the proposition that is uncontroversial for him. To claim this, however, is to provide only the barest outline of a general conception of rationality.[154]

Beliefs in what Foley calls "uncontroversial propositions" are the sort with which my own theory will be most concerned (along with the doxastic practices that result in beliefs in such propositions). Foley's view is presumably that basic beliefs are rational due to their being basic; hence the rationality of such beliefs is spelled out by an account of basic beliefs.

I don't believe it would be appropriate to adopt Foley's theory (even if I thought it were very plausible) for purposes of this dissertation. The theory that I shall advance is specifically designed to aid in solving the problems raised in this dissertation, and Foley's of course is not. There is reason to think that the two theories really concern slightly different, but related, concepts: Foley's concerns the concept that might be loosely described as "having reason to believe." By contrast, mine concerns the concept of "being an instance of the use of reason." Foley focuses on having reasons, while I focus on using reason. These are related but distinguishable concepts.

As we shall see, however, there is one instance where my (rather sketchy) comments might usefully be supplemented by Foley's attempt to characterize the rationality of basic beliefs. But for the outline of the theory I will try to develop a different account, much less detailed than Foley's. I can perhaps be forgiven the lack of detail considering that the purpose of this chapter is, ultimately, to illustrate one possible way get around charges of meta-skepticism, given the results of Chapters 1-3. The theory I shall develop will, I think, suffice for that limited purpose.

So without further ado, I will outline my own theory of rationality.

Various items are called `rational', including actions, habits (physical and mental), people, minds, and beliefs, but it is the rationality of belief and doxastic practices with which we are now concerned. To further complicate matters, there are a variety of ways to approach an account of rational belief; or perhaps it would be better to say that there are a variety of senses in which the word `rational' may be applied to belief.

Two senses will be introduced here.[155] First, there is the sense of `rationality' that is equivalent to an old use of the word `reason' (or, poetically and reverently, `Reason') that denotes certain doxastic practices [156] or habits; second, the results of engaging in those habits, i.e., particular beliefs, can be called `rational'. One may regard the noun, `rationality' or `Reason', as short for the doxastic habits, and the adjective, `rational', as applying primarily to the resulting beliefs. So rational beliefs are the deliverances of the practice of rationality (or of Reason).[157] That this sort of concept of rationality or Reason is robust, there should be little doubt: this notion was common throughout early modern philosophy, indeed being one of the leading epistemic notions of the age. It has perhaps become less common, among professional philosophers anyway, since then.[158] The theory of Reason to be developed presently owes its heaviest debt to one of the greatest of the early moderns, Thomas Reid.[159]

In the sense in question, `rationality' (or `Reason') very generally denotes a habit of believing a certain way in certain circumstances. A bit more precisely, the term denotes a collection (indeed, a highly integrated body) of doxastic practices including those originating with sense-perception, introspection, memory, perhaps a capacity for conception or perhaps some sort of "rational intuition" (which would be the source of a priori knowledge), varieties of reasoning, and reliance on testimony.

Thus, for example, the practice of believing what one sees when fully awake, one's visual apparatus is working properly, and one is in otherwise good circumstances is part of what `rationality' means. The habit of believing that those things one vividly remembers (and has no reason to suspect were somehow fabricated) actually occurred is another part of what `rationality' means.

The "deliverances of Reason" – i.e., the beliefs that result from good doxastic practices – may themselves be described as `rational'. That is, in virtue of the fact that that they are the results of rationality, certain beliefs are rational. However, by contrast, one could instead antecedently identify the rational beliefs and work one's way (not unlike a particularist would have it) to an identification of the rational doxastic practices. Fortunately, it will be unnecessary to discuss any important issues lurking behind that observation.

This theory might be formulated in terms of supervenience or in some other way to avoid talk of meaning, if preferred. One might just as well say that the concept `rationality', in one sense of this word, supervenes on the network of doxastic practices just mentioned. I doubt any crucial point depends on whether the view is stated in terms of meanings, supervenience, necessary truth, explications, or any other such item.

It appears clear enough that beliefs can be rational, in the sense under examination, without being supported by other beliefs. It is not necessarily, at any rate, in virtue of any such support that such beliefs have the property of being rational.[160] For example, suppose I feel a sharp pain in my head, and I immediately form the belief that I am now suffering a headache. The belief is the result of the ordinary functioning cognitive apparatus of introspection, in good circumstances (the headache itself notwithstanding). I am perfectly rational in believing I have a headache, simply by virtue of this belief being the result of an instance of well-functioning introspective processes; and the belief is rational in this sense regardless of whether it is based on or supported by any other belief (except in the tenuous sense that it might be partly based on some "background beliefs").

Obviously, to develop this theory further, it will be necessary to get clear about what practices we're talking about. It is one thing to say that some sense-perceptual doxastic practice is part of what we mean by `rationality'; it is quite another to describe that practice in enough detail to be useful. We will address this problem next.

III. Rationality and Properly Basic Doxastic Practices.

The primary point of obscurity is one of generality. So far all I have said is that rationality is comprised of doxastic practices that originate with sense-perception, introspection, memory, varieties of reasoning, reliance on testimony, and perhaps conception or some sort of "rational intuition." But I do not wish to be understood as meaning that all doxastic practices originating with sense-perception, etc., comprise rationality. I will narrow the concept in two ways.

First, there are some specialized doxastic practices – such as diagnosing illnesses or deciding how to rule on a court case – that, while perfectly rational in a sense, are too specialized to fall into the extension of the term `rationality' as I shall develop it here. It will suffice to develop a "minimal" conception of rationality. Accordingly, I will discuss what I shall call basic doxastic practices.

Second, human beings are of course not always rational in every mental act. There are some irrational doxastic practices, such as forming beliefs about the future based on horoscopes, or concluding that someone is a criminal simply because he wears certain clothes.[161] Hence I shall want speak of properly basic doxastic practices.

So my next tasks are clear: to specify both the conditions under which doxastic practices – are basic and under which they are properly basic. This should provide a clear enough idea of what is meant by `rationality' on this minimal conception.

Some informal remarks will help motivate my approach.

The basic doxastic practices – are those that are fundamental to our interface with the world. They are, moreover, practices the reliability of which cannot be established solely through the use of other practices; one might (very loosely) say that they are self-certifying.[162] Hence, forming simple beliefs about the color of one's hand while looking at it in broad daylight, when one is wide awake, sober, etc., is about as good an example of the use of a basic doxastic practice as one could require. But forming beliefs about the length and color of a stick, when the stick immersed in muddy water, could not be called an example of that same practice. If one can reliably form beliefs about the length and color of the stick in that case, his reliability at doing so can be explained in terms of more reliable and ordinary practices, such as looking at the sticks while out of the water, and then carefully observing how their appearance changes when put into water.

These informal remark provides a clue to how the notion of `basic practice' might be explicated. In short, the basic practices are those such that their reliability can be most plausibly established – if it can be established at all only by means of epistemically circular arguments. It will be helpful to recall the definition of epistemic circularity.[163] If one attempts to argue that basic practices reliably produce a preponderance (or the desired frequency) of true beliefs, then a necessary condition of his being justified in believing either one of the premises, or the connection of the premises with the conclusion, is the truth of the conclusion itself.

An example will be helpful. Consider the following description of a sense-perceptual doxastic practice:

(BSP) A fully-awake, sober, and healthy observer S seems to perceive an ordinary middle-sized object o and its observational property P, in excellent conditions at a close distance; moreover, upon brief reflection S is not aware of any reasons to believe o does not have the properties it appears to have; and on the basis of that apparent perception, S forms the belief that b, where b is the proposition that object o has property P.

For example, I, fully awake, sober, and healthy, seem to see a blue coffee cup on the desk, by all appearances illuminated by the sunshine coming in from the window. I am not aware of any reason to believe there is not a blue coffee cup on the desk. So on the basis of this apparent visual act, I form the belief that there is a blue coffee cup on the desk.

If I were to try to argue for the reliability of BSP (i.e., for the claim that beliefs described by BSP are probably true), it would be an extremely complicated business. But very probably the argument would have to contain at least two elements. First, it would have to contain premises to the effect that at various particular times, I have seemed to perceive various things and various of their properties, in particular circumstances as described in BSP. Second, it would have to contain premises to the effect that at those times, those things in fact did have those properties. Any argument that lacked these elements would, prima facie, be implausible; but an argument that had them could be construed as a track record argument.

But (recall from Chapter 1) how can a track record argument for such a fundamental doxastic practice avoid epistemic circularity? For each of the second variety of premise, some justification is required. For example, when I claim that, as it turns out, there really is a blue coffee cup on my desk, how is my claim supported? Surely by further (perhaps more careful) observations, which are further instances of BSP, no doubt. Prima facie at least, the most plausible sort of argument for the reliability of BSP will be an epistemically circular track record argument.

Perhaps I shall attempt to account for the justification of the premises by means that do not depend on sense-perception. Say I confirm the correctness of my belief via the memory of some testimony (I remember that a minute ago someone told me there is a blue coffee cup on my desk). The direction of this defense is transparently ridiculous, not least due to the fact that testimony requires reliable sense-perception in order to operate. More to the point, it is obvious upon reflection about the means of evidence-gathering we have at our disposal that none of them alone provides the necessary, detailed data that would confirm that what appears to our senses to be the case is in fact the case. No one has ever seriously tried to argue for the reliability of the most obvious examples of sense-perception in this way, and probably no one ever will.

But of course there is a tradition, particularly among empiricists, of trying to demonstrate a close connection between ideas, impressions, sense-data, etc., and external objects. A number of philosophers in this tradition (including, on some interpretations, Berkeley, Hume, Kant, and various logical empiricists of the twentieth century) have suggested that we reduce talk of external objects to talk of some mental objects, thus seemingly obviating the need to demonstrate reliability. But even these philosophers admitted and wished to account for the fact that not all sensory appearances are veridical (however they understood what "veridical" might mean). Regardless of one's view of the existence of things beyond any alleged "veil of perception," the problem of distinguishing reliable sense data from unreliable sense data will remain and with it the problem of epistemic circularity.

But just suppose that an argument for the reliability of BSP avoided circularity: in its premises it either made no reference to instances of apparent visual perception, or made no claims about the truth of beliefs formed thereby. In either case, one could not produce any sort of inductive or abductive argument for the reliability of BSP. So one would have to deduce it, presumably, from other sorts of premises. But it is hard to say, I think, just what those other premises might be. Moreover, to avoid epistemic circularity, they would have to be a priori premises – i.e., premises not supported by sense-perception but instead by some other (alleged) doxastic practice such as rational intuition.

Here the force of the Meta-Regress Argument may again be brought to bear. Supposing that one did produce a successful a priori argument for the reliability of sense-perception (as unlikely as that is, in light both of common sense and of failed attempts to do this by various philosophers [164]), one would still be faced with the task of arguing for the reliability of one's means of acquiring a priori justified beliefs.

There is a specific, limited list of doxastic practices of which remarks similar to the above hold equally well, including those associated with memory, introspection, deduction, and induction. The reliability of the most obviously correct uses of any of these doxastic practices cannot be established purely by means of the others. One cannot establish the reliability of one's most vivid memories, even bringing all the resources of reasoning, sense-perception, introspection, etc., to bear, unless he is permitted to assume, somewhere along the line, that some beliefs formed on the basis of memories are justified.

There are a couple of doxastic practices the status of which is a matter of controversy. Whether testimony could be included on the list has been a matter of some debate, but most recent observers (as well as Reid [165]) believe that the reliability of testimony in general cannot be established without relying upon testimony.[166] Moreover, whether some faculty of a priori intuition should be included – as that by which beliefs in basic truths of arithmetic, logic, etc., gain their justification – rests on (and indeed defines key aspects of) an old debate.

However that might be, I think `basic doxastic practice' can be satisfactorily defined as follows:

Def. A doxastic practice is basic iff the most plausible argument (or series of arguments) for its reliability makes use of the practice itself (and hence displays epistemic circularity ).

I am using the term `basic' for this concept by analogy with the object-level concept of a basic belief. A basic belief is a belief that is justified but not justified by other beliefs; it is a belief that lies at the foundation of a structure, as it were, that is composed of other beliefs. Similarly, a basic doxastic practice is one that is so fundamental that its reliability must be assumed in any attempt to argue for its reliability, regardless of what other doxastic practices might be deployed.

This definition suffers from vagueness in its talk of arguments that are `plausible', but it is, I hope, sufficient to pin down the concept for my purposes.[167] Its vagueness is actually desirable insofar as it makes the decision whether testimony and rational intuition are to be counted basic practices turn on the right questions: do the most plausible arguments for the reliability of testimony or rational intuition display epistemic circularity? Or are all attempts to argue for the reliability of testimony without making use of testimony, or rational intuition without making use of rational intuition, obviously unsound? In any event, the clause `the most plausible' seems necessary to include if only because obviously implausible arguments for the reliability of doxastic practices ought not to be allowed to bear on the decision whether to say the practices are basic or not.

Next observe that, at least as far as the proffered definition of `basic doxastic practice' is concerned, some basic practices might not be reliable at all. In that case, the practices might be basic, but they are not properly basic:

Def. A doxastic practice is properly basic for S iff it is basic and virtually all of its resulting beliefs are among the most obviously correct (compared to all other beliefs).

What are some properly basic doxastic practices? They definitely include forming beliefs based on the most obviously correct uses of – again sense-perception, memory, introspection, and deductive and inductive reasoning, and they probably include testimony, and possibly rational intuition or conception.

The concept of proper basicality for doxastic practices is, by this definition, vague, since it is at least somewhat vague what beliefs are described by `the most obviously correct'. But the point is that the properly basic practices do result in what appear to be, to the vast majority of sane, ordinary adult human beings, obvious truths.

By way of explaining what I mean here, the best I can do (short of writing another dissertation) is give examples and make some vague observations about them. To begin with the examples then: I very vividly remember what my address is, that my fiddle usually hangs on a nail on the wall in my kitchen, that the sun has risen every day, etc. No memories are more obviously correct than these. Similarly, I very plainly see that it is presently day outside, hear that there is music playing, feel myself to be sitting down, etc. There could be no more obviously correct uses of memory and sense-perception than those that resulted in these beliefs.

So this phrase, `the most obviously correct', has a similar function to Chisholm's account of `certainty', which has it that that than which nothing is more justified is, by definition, certain.[168] To put my claim in Chisholm's terminology: the properly basic doxastic practices definitely include (though might not be limited to) those basic practices virtually all of the issuances of which are certain (in Chisholm's sense).[169]

Moreover, among those beliefs that `the most obviously correct' picks out are those that Moore identified in his famous essay "A Defence of Common Sense" – in particular, the common sense beliefs that he said "I know, with certainty, to be true."[170] To go back even farther, the beliefs to which I am referring also include what Reid called "first principles" and "principles of common sense."[171]

Admittedly, there are differences between Moore's and Reid's accounts of common sense (Reid's is far more detailed, for one thing); Reid most usually cites as examples very general claims,[172] whereas Moore cites particular claims. But of common sense beliefs, whether general or particular, Reid and Moore agreed that there are none more obvious than they are. I would be willing to include both varieties of beliefs in the extension of the term `obviously correct'.

I hope these remarks make it clear that `obviously correct' does not mean (literally) true. Someone who objected to my using the term `correct' for what appears correct (to common sense ) might prefer that I use scare quotes as follows:

Def. A doxastic practice is properly basic for S iff it is basic and virtually all of its resulting beliefs are among the most obviously "correct" (compared to all other beliefs).

I have no objection to this.

Beyond the foregoing remarks, it is not my intention to characterize or defend common sense la Reid or Moore but instead to employ their work in motivating a refutation of meta-skepticism. The fundamentals of this philosophy of common sense are of course not universally accepted. My own refutation of meta-skepticism will have to be regarded as conditional, its full demonstration awaiting a more adequate engagement of the topics and objections surrounding the philosophy of common sense.[173]

But I do not yet wish to assert (with Moore, for example) that the most obviously correct uses of our basic doxastic practices result in true beliefs, or even in mostly true beliefs. That's a matter I will leave for a later section in this chapter; for all I have said so far, it might be the case that the most obviously correct uses of our basic doxastic practices result in mostly false beliefs. In that case it would not be particularly impressive to call any practices properly basic.

Here is an example of a basic practice that fails to be properly basic. A common deductive fallacy that may be considered a (narrowly circumscribed) doxastic practice is affirming the consequent: given the conditional `If P, then Q', and given Q, infer that P. (For example: if there is fresh coffee in the cup, the cup is warm; the cup is warm; therefore there is fresh coffee in the cup. But it might be tea this time.) This inference procedure is part of a basic, but not a properly basic, doxastic practice, since any invalid inference rule will fail the test of producing beliefs virtually all of which are "most obviously correct."

Here is a more exotic example of an improperly basic practice. Suppose Helga the Mystical claims to have a faculty of "special seeing." This faculty, Helga says, permits her to "see" events on the other side of the galaxy. Now, there is no way that we could determine if the alleged deliverances of "special seeing" are correct. Moreover, Helga claims to "see" that she has been given this faculty by some mysterious beings on the other side of the galaxy, and no one else has been given this faculty. She offers an argument to the effect that "special seeing" is reliable, on these grounds; but it is, of course, epistemically circular. And indeed, this appears to be the most plausible argument Helga can give. So "special seeing" is a basic doxastic practice, according to the definition offered above. But it is plainly not properly basic, since Helga's pronouncements about alien worlds are not among the most obviously correct she might make.[174]

The above considerations, while still admittedly sketchy, give more substance to the notion of epistemic rationality introduced in the previous section. Rationality, or Reason, on this minimal account, is simply the collection of properly basic doxastic practices. So, first, the practices that constitute (minimal) rationality are limited to (some of) those such that the most plausible arguments for their reliability are epistemically circular: they are "self-certifying," if their reliability can be certified at all (which, on my view, it cannot be). Second, of these, the practices that constitute (minimal) rationality are those such that virtually all of their resulting beliefs are as obviously correct (or "correct") as any beliefs can be.

It will be useful to be able to speak of `standards of rationality', `rationality standards', and `r-standards'. These standards state the relationship between rational doxastic practices, which together constitute (minimal ) Reason, and rational beliefs that result therefrom. An account of such standards falls neatly out of the above work: we simply associate a standard with each identifiable properly basic doxastic practice, such that the practice is described in the antecedent and the belief is described as rational in the consequent.

Here is a quick gloss on the translation scheme. Consider a doxastic practice described in the following form:

The practice of believing that p when such a belief meets (nonepistemic) conditions c (where c might include subjective or nonsubjective states of S and S's environment) at time t, and the belief that p is undefeated.

One may identify an associated r-standard (or several[175]). In general, the standard here could be described as follows:

If S's belief that p meets (nonepistemic) conditions c (where c might include subjective or nonsubjective states of S and S's environment) at time t, and the belief that p is undefeated, then S is rational in believing that p at t.

The converse translation scheme should be obvious. So I will speak of basic standards of rationality and basic r-standards as well as basic practices. A basic standard is simply one such that its associated practice is basic. A properly basic standard is simply a basic standard such that its associated standard is properly basic.

Consequently one may describe minimal rationality as what is described by the antecedents of properly basic r-standards.

IV. Strawson's Dissolution of the Problem of Induction, and Salmon's Criticisms.

A thoughtful critic might concede that the sense of `rationality' developed here does indeed exist, and has been used historically, while insisting that it is not a particularly interesting sense of `rationality'. One clearly interesting sense, by contrast, is the epistemic version of means-end rationality, where the end is truth: in short, we want our beliefs to be rational because forming beliefs rationally is the best means to having true beliefs. But to be told merely that a belief results from some particular uses of sense-perception, for example, is not to be told anything about the chances of that belief's being true. As far as this theory goes, we might get to apply the epithet `rational' to a belief, even though the belief is probably false. And why should we care at all that a belief is rational in this sense?

This is very similar to a criticism that Wesley Salmon made of a solution to the problem of induction credited to P.F. Strawson. Their exchange will be instructive to review now; from it, we shall be able to draw some important lessons about how the theory must be further developed. So let's begin with Strawson.

Strawson formulated a solution to the problem of induction, or perhaps it should be called a "dissolution" of the problem, by analogy to a point he makes about deduction.[176] Strawson advanced the view that to ask seriously whether deduction in general is valid would be "senseless," since "to say that an argument, or a form or method of argument, was valid or invalid would imply that it was deductive; the concepts of validity and invalidity had application only to individual deductive arguments or forms of deductive argument."[177]

In a similar way, Strawson said, as far as inductive arguments go, to be "reasonable" in making them is no more or less than to follow (or "apply") inductive standards:

Similarly, if a man asked what grounds there were for thinking it reasonable to hold beliefs arrived at inductively, one might at first answer that there were good and bad inductive arguments, that sometimes it was reasonable to hold a belief arrived at inductively and sometimes it was not. If he, too, said that his question had been misunderstood, that he wanted to know whether induction in general was a reasonable method of inference, then we might well think his question senseless in the same way as the question whether deduction is in general valid; for to call a particular belief reasonable or unreasonable is to apply inductive standards, just as to call a particular argument valid or invalid is to apply deductive standards.[178]

We can distinguish good from bad inductive arguments, of course, or reasonable from unreasonable, and it is well-known that the task of drawing such a distinction is daunting. But if the question is put whether induction in general is reasonable, Strawson says, the question is senseless: following inductive standards is part of what being reasonable is. Part of rationality (no small part either, I might add) consists of the habit of following inductive standards.

Some pages further on, Strawson put his point in terms of analyticity and meaning:

It is an analytic proposition that it is reasonable to have a degree of belief in a statement which is proportional to the strength of the evidence in its favour; and it is an analytic proposition, though not a proposition of mathematics, that, other things being equal, the evidence for a generalization is strong in proportion as the number of favourable instances, and the variety of circumstances in which they have been found, is great. So to ask whether it is reasonable to place reliance on inductive procedures is like asking whether it is reasonable to proportion the degree of one's convictions to the strength of the evidence. Doing this is what `being reasonable' means in such a context.[179]

Writing in a more innocent pre-Quinean-"Two Dogmas" era, Strawson had few qualms about making his point in terms of analyticity and meaning; but again, the point need not be made that way. One could just as well say, for example, that the rationality (or reasonableness) of inductions supervenes on the adherence to certain canonical standards of induction; or that, necessarily, if an induction is in accordance with some canonical standards of induction, then it is rational.

In any event, Strawson was very aware that this sort of solution to the problem of induction is not likely to be very satisfying to some, so he labored to make it more intuitive; but it will not add to our present purpose to review the latter work here.[180]

Strawson's solution to the problem of induction is similar to my own theory of rationality: a belief can be rational, in the minimal sense developed, by its being a result of one of the doxastic practices (of which following inductive standards is one) that together constitute rationality. So my theory enlists the Strawsonian defense of the validity of deduction in general, and of the reasonableness of induction in general, and devises a similar Strawsonian (not to mention Reidian ) account of the rationality of basic beliefs formed on the basis of our other ways of learning about the world: perception, memory, and the rest.

Salmon had a number of different criticisms of Strawson's view,[181] that can without trouble be converted into criticisms of my view of rationality. Let us take the criticisms in turn.

First objection. Salmon declares, after briefly describing a theory about induction that he ascribes to Strawson, "If the foregoing theory is correct, empirical knowledge is, at bottom, a matter of convention. We choose, quite arbitrarily it would seem, some basic canons of induction; there is no possibility of justifying the choice."[182]

This is a non sequitur, although it might have had more force with readers when Salmon was writing (in 1957). That it is a non sequitur should be obvious because it is obvious that there are other possibilities, that account for the origin of inductive standards, in addition to an arbitrary choice that determines a convention. And Strawson's position does not commit him to any of these possibilities in particular. There is nothing about the claim that inductive reasoning is part of what we mean by `rationality' that commits Strawson to the view that the standards of induction (and thus empirical knowledge) are a matter of convention.

One might just as well say that it is purely a matter of convention that water is H2O (since chemists happens to define `water' that way), or that the bachelors of the world happen not to be married at present. But the only item along these lines that is a matter of convention is the connection of the word with the concept. The word `rationality', as distinguished from most other words, has been arbitrarily associated in English with a certain collection of cognitive practices (that includes induction ); that's granted. But surely Salmon would not wish to infer from this truism that it is a matter of convention that the practices themselves are rational. But that does appear to be the inference behind Salmon's objection to Strawson's view.

Indeed, among the possible origins of various standards of induction, the one that leaps to mind as most plausible is not arbitrary convention but instead some manner of inbred, natural tendency developed no doubt through some evolutionary processes. As Quine famously observed, those who are poor at induction have a "pathetic but praiseworthy" tendency to die off. It is perfectly consistent with Strawson's view (that following inductive standards is part of what we mean by `reasonable') that inductive standards should have an evolutionary origin. Beyond the simplest sorts of induction, however, something like the Method of Reflective Equilibrium will generate a canon of inductive standards (as I will explain in Section VII below).

Second objection. Here is an objection, commonly raised,[183] to Strawson's sort of move. Suppose that I advance the views that some inductive doxastic practices – are rational because the practices constitute rationality and that inductively formed beliefs are perforce rational as well. Then, the objection goes, I still haven't solved the problem of induction, because I haven't identified which of the many possible inductive rules are the correct rules to follow. Perhaps I can say that some inductive practices are rational, but I haven't specified which ones they are. Indeed, as Goodman insists,[184] there really isn't any more to the problem of induction than the challenge of giving a precise account of correct inductive inference.

My theory of rationality provides only that some inductive doxastic practices – are rational, and I haven't said which ones. So one might be prepared to grant my claim, while maintaining (rather dubiously) that it is not controversial or at least not helpful. To distinguish the good inductive practices from the bad ones, one must have some way to make the distinction. How will I make this distinction without relying on some inductive rules (either to confirm some particular inductions, or to derive the rules that I say are good) and thus committing the very epistemic circularity I am laboring to avoid? In short, while I might have established that engaging in some generally-described doxastic practices is rational, I have not solved a problem that is just as crucial, viz., how to decide which of competing specific practices that fit in the general description (not all of which are rational) are the best.

Essentially, the challenge is to make the definition of `rationality' precise by specifying very clearly which sorts of belief are "minimally rational" on account of their being formed as the result of (in this example) an inductive inference.

My response to this objection comes in two parts.

First, I may simply concede that, indeed, we lack a precise account of correct inductive inference and, by extension, of rational belief or Reason. But so what? Most of us are able to go through our lives, even those relatively rare moments in our intellectual lives in which we specifically claim to have rational beliefs, embarrassingly inept at specifying the conditions when our beliefs are indeed rational. In spite of that, we can and in some cases do judge that many instances of belief are as obviously correct as they could be, and that they are rational. As is the case with so many of our concepts, we don't seem to be any worse off for lacking a a precise account to cover these and other cases of our rational belief. So one central claim of the objection – "to distinguish the good inductive practices from the bad ones, one must have some way to make the distinction" – is clearly false, in many cases anyway. In many cases, it is quite plausible to say that not only do we not need a precise account of rationality to be rational, we also do not need one in order to know that we are being rational.

If our concern is to avoid skepticism, then we will be concerned to solve the justificatory problem of induction. Strawson's view (and mine) suffices to satisfy that concern. But if our concern is to get an exact idea of that wherein rationality consists, then we will have to engage in what has long been known to be a very complicated business.

In the second part of my response, which follows, I will attempt to address this latter concern. In discussions of the problem of induction and of similar problems associated with the justification of deduction and other topics, something like the Method of Reflective Equilibrium has been repeatedly proposed as a way to generate more precisified versions of unclear, general formulations of rules.[185] As we saw in Chapter 3, this gives meta-coherentism no small amount of attractiveness; to apply the Method of Reflective Equilibrium to j-standards is closely and naturally associated with meta-coherentism. But the problem with meta-coherentism, as I argued in Chapter 3, is that it is committed to the benignity of epistemic circularity while offering no unique resources to defend epistemic circularity against charges of viciousness.

Still, the same insight can be appropriated to develop the present theory of rationality, and this can be done without committing its defender to meta-coherentism. Essentially, the proposal is to develop and at the same time justify a theory of correct inductive inference by starting modestly. One finds sufficient conditions for weak inductive support by testing an account of such conditions against the least controversial (commonsensical) examples of excellent inductions. In so doing one applies, and takes for granted, simple, modest inductive standards. The following might be serviceable as a "simple, modest inductive standard" that one could (perhaps after some adjustment) take for granted in developing a modest theory of correct inductive inference:

For any weak account of inductive inference A, according to which an inductive inference has some inductive presumption in its favor, A itself has some presumption in its favor if A provides a model of a very wide variety of the most obvious examples of correct inductive inferences.

Obviously, the foregoing is not a principle of common sense in the sense that it just immediately springs to anyone's mind. Indeed, it probably needs more adjustment itself. But it (or something like it) is among the most obviously "correct" things that one can say about which inductive inferences have some sort of support, and it is also designed to apply to accounts of inductive inference.

After taking some such basic inductive standards for granted, one develops and supports stronger and more sophisticated versions of the theory, which applies to a greater variety of inductive inferences. This neither involves one in an epistemic circularity, nor does it commit one to meta-coherentism.

More discussion of this sort of proposal may be found in Section VII below.

Third objection. Suppose that we arrive at some canonical view about what a good induction is (which is not merely conventional, and which was arrived at via something like reflective equilibrium ). The full force of a third objection from Salmon could still be felt. Salmon writes:

It sounds very much as if the whole argument (that reasonable beliefs are, by definition, beliefs which are inductively supported) has the function of transferring to the word "inductive" all of the honorific connotations of the word "reasonable," quite apart from whether induction is good for anything. The resulting justification of induction amounts to this: If you use inductive procedures you can call yourself "reasonable" – and isn't that nice! (p. 42)

More to the point, as I said at the beginning of the present section, why should we care about a conception of rationality according to which it might be true that a rational belief is extremely improbable? The reason we want rational, or reasonable, inductive conclusions (or any other kind of beliefs) is that we want the truth.

It is for this reason that the accounts of `rationality' and `rational' given (or alluded to) above are admittedly not adequate by themselves. An account of the relationship between reason and truth must also be developed in order to show that indeed the sense of `rationality' in question is worth caring about – that this sort of rationality is indeed something that, if possessed by a belief, could provide justificatory support to other beliefs. Specifically, to do the job it is required to do here, rationality must be truth-linked. In the next section, I will turn to this refinement.

V. The Principle of Rationality.

Why should we care about a conception of rationality according to which it might be true that a rational belief is extremely improbable? Who cares about rationality if it doesn't elicit the (probable) truth?

Here's one possible answer: truth, and probable truth, are themselves to be assessed according to the same methods that we are honoring with the name `rationality'. So it is not surprising that rationality should elicit truth, since we determine both rationality and truth in just the same ways.

Of the many things one might say in reply to this claim,[186] one serves to focus the discussion most sharply: viz., the claim doesn't address the question. The question is, essentially, "Are rational beliefs probably true, in the sense of `rationality' in question, and if not, who cares about rationality?" It might indeed be correct to say that both the rationality and the truth of a belief are to be determined in the same ways. But that only says how truth is determined; it doesn't answer the question whether rationality does in fact elicit the truth.

If one treats these two questions as synonymous, one assumes that there is no more to truth than what we can determine the truth to be, i.e., one assumes the correctness of the pragmatist/anti-realist program. The skeptic might be happy to admit that truth and rationality are determined in the same ways: it's just that we don't know if our ways of determining the truth actually do give us the truth. To simplify, the pragmatist disagrees with that, holding that truth is what we can determine it to be, and thus essentially rules some kinds of skepticism out of court without a hearing.

The reply that I shall advance does, perhaps, treat the skeptic in an equally peremptory way, but it does so while maintaining a distinction between truth and what-we-can-determine-to-be-truth (warranted assertibility, pragmatic truth, etc.).[187] I shall simply assume, without argument of any kind at all, that rationality elicits truth. In fact, I would like to dignify this assumption by giving it the grand title The Principle of Rationality , or PoR:

Beliefs formed as a result of those doxastic practices that constitute rationality are, probably, true.

More briefly, rational beliefs are probably true.[188] I believe this principle, added to the account of rationality introduced in Section II, is the most promising way to rescue Strawson's defense of induction from Salmon's objection (the second one spelled out above). When Salmon jokes, "If you use inductive procedures you can call yourself `reasonable' – and isn't that nice!" the proper reply is: "Yes, it is, because it means many of the beliefs I have formed by induction are probably true."[189] And this implication is something that I assume without argument.

Let me try to be a bit clearer about what is being assumed. Since `rationality' in the sense in question is short for a list of doxastic practices including sense-perception, memory, reasoning, etc., the PoR is equivalent to saying that sense-perception, memory, reasoning, etc., are (at least in their most basic uses) reliable. For a doxastic practice to be reliable is just for that practice probably to result in true beliefs. So one could just as well say that I assume, without argument, that the best doxastic practices associated with sense-perception, memory, reasoning, etc., are reliable. There is no need to go through complicated twistings and turnings in trying to avoid epistemic circularity, as the targets of Alston's Reliability of Sense Perception do; the reliability of sense-perception, and of other leading doxastic practices as well, can be assumed without argument.

I should also clarify what I mean when I say I assume the PoR. In claiming this I mean that I accept the Principle but I make no attempt to argue for it. It is not part of my meaning (though this is a closely related issue) that my belief is not justified (i.e., it lacks that the property of justification); nor is it part of my meaning that my belief is not rational. It so happens (as will be made clear in the next section) that I hold that my belief in the Principle is not justified (nor is it unjustified; it is nonjustified ). I also believe that my belief in the Principle is rational, but I am not concerned to argue for this and no part of this chapter's argument depends on the claim that my belief in the Principle is rational.

I have no doubt that some people will be thoroughly unimpressed with this move – although, as we shall see in Section VIII below, Alston [190] is (or at least was) willing to accept something like this solution. In any event, I have laid out a serious problem, to wit, that rationality in the sense introduced here might have little to do with truth, and solved it by "assuming it away." More importantly, the assumption I propose to make would dismiss, peremptorily it appears, the Problem of Meta-Justification and the meta-skepticism that looms behind it.[191] Why indeed should anyone take such a move seriously?

Most of the rest of this section will be devoted to answering this question: "Why should anyone take the mere assumption that rationality elicits truth seriously?" For anyone to take the assumption seriously, he would have to suppose that I am, or at least I could be, within my rights to assume it. And to declare (as I do) that I am within my rights to assume the PoR without argument is to presuppose I have no burden to prove it. So the question now at issue is closely related to some other questions: "Why think that I am within my rights to assume what I want to assume?" and "Why think it at all plausible that there is no burden to prove what I want to assume?"

Suppose I did succeed in showing that I am "within my rights" to assume the PoR. This makes it sound as though I want to say I am justified (on a deontological account of justification) in assuming the Principle. Is that correct?

It is not. In claiming that I am "within my rights" to assume the PoR, I do not wish to claim that I am epistemically justified in making the assumption; I mean, rather, that I am not flouting any ordinary standards of philosophical discourse (of the same sort as those I mentioned in Chapter 2, Section IV).

In that case – a critic might wonder precisely how does the claim that I am not flouting "ordinary standards of philosophical discourse" in assuming the Principle differ in meaning from the claim that I am justified in accepting the Principle itself? In other words, if I do succeed in showing that I am within my rights to assume the Principle, will I not have thereby demonstrated my justification in accepting it? And in that case, why am I saying I merely assume it?

To answer these questions, I will describe in more detail just what standards of philosophical discourse I am claiming not to flout.[192] It should be self-evident to anyone familiar with ordinary accounts of justified belief that the satisfaction of such standards is neither necessary nor sufficient for justified belief, on such accounts.

As a general rule, one has a burden to support whatever claim he makes. I will call this the "Burden-of-Proof Rule." But there are certain exceptions to this rule; I will cite four.

The first exception can be expressed as follows: if it is simply impossible to give a non-question-begging argument for a claim, and the claim is as obvious as any claim can be, then one has no obligation to prove it. Examples of such a claim are some first-person subjective reports, e.g., claims to having sharp pains or to having the impression of seeing certain colors, etc.[193] If it is impossible to argue for an obvious claim, one is not obligated to argue for it; ought implies can, so impossibility implies lack of obligation.[194] Of course, one must bear in mind that it might be quite controversial whether a given claim is amenable to direct argument, and also whether it is "obvious" or not.

The second and third exceptions are due to pragmatic considerations – i.e., considerations surrounding the context, particularly the shared assumptions, of discourse.

The second exception to the Burden-of-Proof Rule rests on the observation that there are reasons, after all, that the Burden-of-Proof Rule is so widely urged. One who insists on its observance must, unless his insistence is frivolous, believe that there is some way that the (alleged) burden of proof can be met – not necessarily on behalf of a particular claim (which might, after all, be indefensible), but in general. So the very use of the Burden-of-Proof Rule carries with it certain pragmatic assumptions, e.g., that language exists, that the language used has enough agreed-upon meaning for meaningful communication to be at least possible, that something that can in some cases serve as "proof" is possible, etc. If these propositions were not true, the very procedure (or language game, to speak Wittgensteinian) of asking for and supplying proof would be pointless. So it is reasonable to exempt such assumptions from the Burden-of-Proof Rule itself.

A third exception to the Burden-of-Proof Rule occurs when a similar sort of pragmatic argument is available. Suppose there is a claim that satisfies the following conditions. Either this claim, which is allegedly in need of support, is true or not; if it is not true, then (for whatever reason) there is no point in continuing the discourse. Now, for all that, it might be correct that there is no point in continuing the discourse, because the claim in question is false. But given that, in any case, we want to pretend at least that there is a point to continuing the discourse, it is acceptable to assume the claim in question.

A fourth exception to the Burden-of-Proof Rule occurs when all parties to the discourse in question agree that the claim in question is true. There might still be some interest on the part of some parties as to why the claim is true, i.e., how it might be supported, e.g., against a fictional skeptic. But for purposes of discourse, agreed-upon claims are commonly regarded as not in need of support – and all the more so if it is psychologically impossible for any of the parties to the debate to disbelieve the claim in question.

It so happens that my belief in the PoR is a case in which all four of the above exceptions apply, as I shall now try to show.

1. The assumption is not being made hastily. I propose essentially to "assume away" the Problem of Meta-Justification. But this does not mean that the problem has been dismissed in a hasty, peremptory fashion. After all, we have carefully framed the problem and examined and tested, in considerable depth, a variety of solutions to it. It is only after all of that work that the present approach to the problem is being proposed.

Recall that the problem involves asking the question, "For some standard of justification, how can we justify it?" and then observing that, in general, justifications offered for standards are themselves licensed by further standards. That generates the familiar regress and the problem of epistemic circularity. After the work of Chapters 2 and 3, upon a re-examination of the problem, one might very well conclude that it is an unsolvable conundrum, a problem that cannot be solved directly, in which case it should not be surprising that the proper response to the problem is to "dissolve" it.

In this connection, recall the first exception to the Burden-of-Proof Rule: if it is simply impossible to give a non-question-begging argument for a claim, and the claim is as obvious as any claim can be, then one has no obligation to prove it. A solution to the PMJ is tantamount to giving a direct argument for the PoR. I believe we have found that such an argument is impossible. But the Principle is indeed very obvious – not, perhaps, as obvious as any can be, but quite obvious. Hence I do not believe I have an obligation to prove it. This approach begs the question only in a relatively weak sense: it assumes something essential to what was originally up for proof.

Of course, the mere fact that the problem has appeared intractable so far does not, by itself, give one the right to make what might appear to be an offensively question-begging response, and assert that the statement of the problem "rests on a confusion." Indeed, it only makes such a response a bit more plausible. Fortunately, in the present case, more can be said in defense of such a response.

2. It is pointless to use Reason to establish the reliability of Reason. It should come as no big surprise that epistemic circularity faces us if we undertake a frontal attack on the Problem of Meta-Justification. Here is why. Such an attack essentially involves, for any doxastic practice, an attempt to argue that it is reliable. Of course, any attempt to argue for its reliability will make use either of premises (beliefs) outputted from the very doxastic practice in question or of some other. But assuming that one is approaching the problem in a rational way, all of the doxastic practices one might use make up what we call "rationality." Hence one is attempting to use rationality to prove that rationality is reliable. Or, more precisely, one uses a part of rationality to prove that another (possibly the same) part is reliable; in that case the reliability of the first part remains to be established, presumably by rational means as well. This general procedure is pointless.

I say the attempt to demonstrate the reliability of our basic doxastic practices – amounts to an attempt to use Reason to demonstrate that Reason will give us the truth. In that case, we should simply explicitly assume that Reason does give us the truth, i.e., just take the PoR for granted. After all, if we are asked to give some reasons to believe the PoR, presumably the request is made because the skeptic thinks that this giving of reasons will make the PoR more probably true. And this assumes, of course, that the sorts of considerations adduced by acceptable argument will serve to increase the PoR's probability. What sorts of considerations are acceptable? If any others are, then indeed those described as `rational' according to our theory of rationality are, it being a minimal account of rationality.

The skeptic is apt to complain that here I am saddling him with a concept of acceptable argument he is not necessarily prepared to accept. But if not this concept (which concerns merely a very modest subset what should be considered rational beliefs), what then is acceptable? If the skeptic is attempting to keep us honest – to make us satisfy our burden of proof – then he himself does, pragmatically, take for granted some concept of rational, or otherwise epistemically upstanding, belief. And then what premises are epistemically upstanding?

Perhaps the skeptic will reply that the request for argumentative support does not commit him to any particular notion of acceptable belief; he requires only some sort of (noncircular) argumentative support, not the use any particular kind of premises.

Such a reply would be disingenuous, unless the skeptic were playing some sort of patently pointless academic game, the object of which is to require arguments from all comers and then supply criticisms of all premises. For the Burden-of-Proof Rule to be applied with a purpose, there must be a robust concept of how the burden of proof can be met; that involves, among other things, some tolerably clear notion of which sorts of beliefs are acceptable. Our minimal theory of rationality is (among other things) an attempt to specify a small subset of such beliefs.

It is reasonable, then, to suppose that the non-frivolous skeptic is pragmatically committed to the acceptability (to some degree) of beliefs designated as `rational' according to our theory. In other words, someone who applies the Burden-of-Proof Rule to the PoR with a serious, nonfrivolous purpose is pragmatically committed to the PoR. Thus he is asking us (perhaps among other things) to use minimally rational beliefs to support the claim that minimally rational beliefs are probably true. What makes the request non-frivolous is precisely our shared assumption that those sorts of belief are indeed probably true. But if we have that shared assumption, there is no point to making the request. We may as well explicitly take the PoR for granted.

Note again that this is not meant to be an argument, either direct or indirect, for the PoR. It is meant to be an argument that there is no burden to prove the PoR.

3. A Reidian pragmatic argument. This one can be stated more briefly. Either the PoR is correct or it is not. If it is not, then one (or more) of the basic ways we have of knowing about the world is not reliable, in which case, very probably, this and all reasonings are pointless anyway, being based on bad information. The other possibility is that the PoR is correct.[195] In either case, it is unreasonable to insist that there is a burden to prove it. If the principle is correct, we haven't gone wrong in assuming it; the reason we would want an argument supporting the PoR, after all, is in order to show that it is probably correct. If it is incorrect, the falsehood of the principle renders any consideration of dialectical rights pointless anyway, because, essentially, anything goes. In either case, there is no point in saddling us with a burden to support the PoR.

Again, note that the conclusion here is not that the PoR is correct, but rather that the Burden-of-Proof Rule does not apply to it.

I think the only plausible reply to this argument is to maintain that our account of rational belief might have the precise account of rational belief wrong. In that case, the PoR might be wrong, but we would still have some other, more rigorous notion of rationality to apply. The question then is how we can decide which of competing accounts of rationality are correct. This is, of course, one of Salmon's objections to Strawson all over again. To reiterate what I said about that objection: evidently we must follow something like the Method of Reflective Equilibrium in formulating just how we want our account of rationality to read. The sketch of rationality given in this chapter should suffice for present purposes; no account of rationality that can seriously aspire to the name will contradict this view, though it might greatly clarify or add to it.

4. Acceptance of the Principle of Rationality is virtually universal and involuntary . Consider the claim being taken for granted. It is entirely natural and overwhelmingly widespread to assume, without argument (and even without conscious thought) that the beliefs described as `rational' by our theory of rationality are correct. We take this for granted throughout our everyday and scientific endeavors: "Beliefs formed as a result of those doxastic practices that constitute rationality are, probably, true."[196]

Of course, throughout the bulk of human history and presumably human prehistory, more fully developed views of rationality were not understood or accepted. And the principle, as formulated, would likely be rejected by a variety of irrationalists, for religious, political, or other reasons. So how can I say that the principle is something accepted nearly universally?

It has taken humankind a long time to arrive at the present Western scientific conception of that wherein rationality consists. E.g., sophisticated standards of induction and probability theory, which in modern Western civilization are taught, accepted, and practiced as a matter of routine (by sophisticated thinkers) first had to be formulated and promulgated. Similarly, deductive logic did not spring full-blown from the heads of early barbarian hordes, or any other peoples; it had to be developed by Aristotle, Leibniz, Frege, and many others.

Indeed, virtually all of the very basic ways we have of knowing about the world are, in various professional and academic contexts, subject to carefully-formulated standards. Scientists, physicians, and lawyers have very specialized, stringent standards governing what should count as a reliable observation. Cognitive psychologists study memory with a view to informing historical interpretation and courtroom decisions based on eyewitness testimony, and lawyers, journalists, and researchers of all sorts have elaborate standards governing the reliability of testimony. Experimental psychologists also caution against easy conclusions about the operation of the human mind drawn from introspection – a point that, once heeded, revolutionized psychology in the twentieth century.

All of this is simply to say that we do have sophisticated standards of justified belief. But those same standards have elaborated the modern, scientific, and professional conception of what is rational and what is not. (And a large part of a modern liberal and scientific education consists in being trained to understand and employ that conception; at some colleges today, over $20,000 per year is required to make a student fully rational.)

It is this detailed, sophisticated, highly-developed conception of rationality that is rejected (or parts of it are rejected) by irrationalists of various stripes. At least some religious fideists and mystics might be said to reject the view that reasoning (and particularly arguments about the existence and nature of God) can elicit the truth. Superstitious folk of all sorts reject the claim that in order to know the nature and cause of (at least some sorts of) empirical objects, experiment or careful observation by the ordinary senses is necessary. And postmodernists, radical academic feminists, and others in contemporary academia self-consciously reject the necessity of careful, rigorous reasoning for many topics, saying that whatever knowledge is possible can be had in other ways (e.g., via something like feminine intuition, or perhaps just toeing the current party line).

This being admitted, virtually no one, including such irrationalists, unless they are literally insane, ever honestly denies that rationality produces the probable truth, in a more basic, fundamental sense of `rationality', the sense in which "Man is the rational animal" is true.[197] Aristotle's dictum does not mean that man is the animal that can properly depose a witness, prove theorems in modal logic, apply the laws of the probability calculus, and make accurate descriptions of what he sees on a microscope slide. Our daily dependence on (and facility with) all manner of commonplace generalizations, observations, the simplest of inferences, and so forth is adequate demonstration of this fact. It is rationality in this humble sense that so obviously elicits truth.

And as I shall argue below, we have no choice in the matters of whether to believe that these "most obviously reliable uses of the practices that constitute rationality" are reliable; nor do we have any choice whether to believe some particular issuances of these practices. We cannot, for example, decide simply to disbelieve what our eyes tell us and what we vividly recall. This is not only rational, it is involuntary. Indeed, it appears that the extent to which beliefs about doxastic practices – and their results are involuntary (for the vast majority of humankind, at least) is the extent to which those practices can be described as `rational' in the sense in question. In other words, one may say that man is the rational animal at the very least to the extent to which he has no choice but to rely on certain doxastic practices, which practices constitute a minimal sort of rationality.[198]

Recall now the fourth exception to the Burden-of-Proof Rule: a proposition that is accepted by all parties to a dispute need not be supported. That's precisely the case here. So it would appear we have no burden to prove the PoR.

There is a possible difficulty in applying this exception to the rule to the PoR, namely, that, regardless of whether in fact the PoR is accepted by all parties to a dispute, some of them might deny accepting it. I am not convinced that the fact that a skeptic claims to deny or withhold the PoR should be allowed to make any difference for purposes of assessing whether one really does have a burden of proof, but I am willing to concede that the issue is unclear. Nonetheless, even if this exception were objected to by skeptics who claim to deny or withhold the PoR, there would still be the three other exceptions to trot out.

Let's review this section's four points. First, the assumption of the correctness of the PoR was made only after carefully examining the alternatives. Second, it appears pointless to use rational doxastic practices in order to argue for the reliability of rational doxastic practices. Third, either the PoR is correct or it is not; in either case, no harm is done by assuming it without argument. And fourth, after all, the PoR is something that, properly understood, we all have no choice but to believe anyway.

These points are enough, I think, to discharge my burden to prove that I have no burden to prove the PoR. But they are not, nor are they intended to be, a defense of the PoR itself.

VI. Nonjustified Beliefs.

The foregoing theory of rationality was outlined specifically in service of a solution to a problem, and it will be helpful at this point to review the problem in order to understand what tasks still lie before us.

I said that Chapters 1-3 supported the following conclusion:

(C) For any given justification standard, J, acceptance of it is either not justified, or ultimately receives support only from beliefs that are themselves not justified (regardless of how many intervening beliefs there might be between the ultimately supporting beliefs and the standard).

This disjunctive conclusion appeared to entail meta-skepticism. But in an attempt to avoid a meta-skeptical conclusion, I proposed that some nonjustified but rational belief might license arguments for j-standards, thereby plausibly avoiding meta-skepticism while satisfying the second disjunct of (C): j-standards ultimately would receive support from rational beliefs that are themselves not justified.

Two tasks are before us now: first, to argue that indeed the required supporting beliefs are "nonjustified but rational"; and, second, to explain the solution in more detail. The present section is devoted to the first task, and the next is devoted to the second.

We call beliefs that are neither justified nor unjustified, because the concept `justification' is inappropriately applied to them, nonjustified . It is not contentious to claim that unjustified beliefs exist, but it might well be contentious to claim that nonjustified beliefs exist. So why suppose that there are any nonjustified beliefs?

To summarize in advance (and oversimplify): there are some beliefs that it is psychologically impossible for us (normal adults) to withhold and, thus, for which we are not responsible; so we have no particular permissions or obligations with regard to such beliefs, since permissions and obligations attach only to acts for which we are (at least possibly) responsible; but since justification is a deontological notion, it makes little sense to say that those most basic beliefs that we cannot withhold are justified (or unjustified). I will develop this argument in four parts.

This line of argument bears some resemblance to that of William P. Alston in his essay, "The Deontological Conception of Epistemic Justification."[199] Alston's conclusion is as follows:

We have examined several forms of a deontological conception of epistemic justification in terms of freedom from blame in taking up a certain propositional attitude. All of these but one was seen to be untenable by reason of requiring a degree of control over our propositional attitudes that we do not enjoy. The only version that escapes this fate was seen to be not the sort of concept we need to play a central role in epistemology. Therefore, despite the connotations of the term, we are ill advised to think of epistemic justification in terms of freedom from blame for believing.[200]

The following, however, does assume that justification is a deontological concept, which is something I take to be a matter of semantic fact (the admitted "connotations of the term" at the very least, as Alston says). Plantinga, in Warrant: The Current Debate,[201] unlike Alston in the above-mentioned essay, takes this to be one main reason to demote justification from its privileged place. The reader may if he wishes construe the present section (and by extension, this entire dissertation) as another argument for that thesis.

1. It is psychologically impossible to withhold some particular kinds of belief. I maintain that there are some propositions that it is psychologically impossible to disbelieve or even withhold (at a given time). For example, it is impossible for me (in present circumstances, of course) to disbelieve or withhold the belief that my name is "Larry Sanger," that I have been alive for longer than a day, that as I write it is day and the sun is shining, that 2+2=4, that other persons exist, etc. There is a potentially infinite list of such platitudes that I could not so much as withhold. In some sense,[202] in present circumstances, I must believe them. I will describe such beliefs as `irresistible'.

That many of our beliefs are not under our immediate voluntary control is a common view among contemporary epistemologists, who (elaborating the view) comment frequently that beliefs of the sort just listed certainly cannot be denied, at least not immediately and without great effort (and perhaps for most people, at all).[203] This point is emphasized by Wittgenstein at length.[204] The point is simple enough when properly understood, but some clarifications are in order.

As a philosopher, I might write as though I could doubt this sort of obvious truth (e.g., when discussing Cartesian skepticism, I might pretend to doubt that I possess legs). But it does not follow from the mere fact that I and others can discuss such doubts at length (indeed, with considerable erudition and footnotes) that we can actually succeed in making ourselves have them.

Those contemporary philosophers fond of pointing out that we lack privileged access to our mental states should be comfortable admitting this. But before misgivings about privileged access became popular, Hume (famously) admitted that he could maintain his skeptical doubts but only in the confines of his study,[205] which admission has inspired some interesting reflection on the nature and possibility of being a skeptic. My present claim, entailing for instance that Hume could not doubt the existence of the desk at which he philosophizes, seems to be at odds with his own claim to doubt this very sort of thing (while in his study). My claim could be construed as an attack on Hume's skepticism;[206] but it needn't be. I can at least concede the point to the skeptic (without unnecessarily engaging this particular issue at length) that we can understand doubts and skeptical propositions and debate their merits, even if at no point do we ever succeed in believing them. And such debate can, of course, be of tremendous value.[207]

Surely some beliefs that one could not at a given time so much as withhold can and do change nonetheless. We can conceive of circumstances in which I discover that what I firmly believed to be my name is not my name, that it is night when it appears to be day, and so forth. So to say that some particular beliefs are presently irresistible is not to hold that, through the presentation of the right kinds of evidence, we could not (in any circumstances) eventually be persuaded otherwise. Beliefs that I claim to be presently irresistible are not, for all that, infallible.

Next I wish to identify a class of beliefs that are irresistible in the above-described sense. For very many adult human beings, very often, the belief that their own senses supply them with basically accurate information cannot be denied. Similarly, at most times most people would find it impossible to doubt that their most vivid memories of events are of events that actually occurred (even if they would admit that they could be wrong about some details). The same remarks apply to other of the most basic ways we have of gathering information about the world: introspection, testimony, rational intuition and/or concept-formation, deduction, and induction.[208]

In general, then, I would endorse the principle that, at most times, most adult people cannot (if they try, for whatever reason they might do so) withhold beliefs resulting from the "most obviously correct" uses of sense-perception, introspection, memory, etc. This principle is, I think, probably true in stronger versions; i.e., I think that the vast majority of people cannot, even if they try, doubt the deliverances even of less-than-most-obviously-correct uses of sense-perception, introspection, memory, etc. But we can employ the principle in its relatively weak version.

Distinguish between individual irresistible beliefs that result from these doxastic practices, and second-order beliefs, equally irresistible, that these doxastic practices are reliable. I think it is fairly obvious that both classes of belief are, for most persons at most times, indeed irresistible.[209]

2. We are not responsible for these beliefs. The next point can be developed more briefly. It is a platitude that normative responsibility [210] for an act (that is, praiseworthiness, culpability, blameworthiness; as distinguished from merely being a cause) requires freedom, in some sense. If I am responsible for stealing, then I stole freely. So if I am not free to act other than how I do act, then I am not responsible for the act I perform (even if I caused it to happen). There are ancient disputes to be settled about what `free' means in these sorts of locutions, but nothing I shall say here will depend on any particular theory of freedom.

It is an equally obvious platitude that, if I cannot do otherwise than what I in fact do, then I am not acting freely. A traditional (and notoriously vague) account of `freedom' (better viewed as a constraint on more detailed theories of freedom) has it that I am free if I could have done otherwise.

Consider now these two platitudes together. It follows from them, together with the observation in the first part of the argument above, that we (the vast majority of adult human beings) are not responsible, in a normative sense, for holding the belief that the most obviously correct uses of sense-perception, introspection, memory, testimony, rational intuition and/or concept-formation, and reasoning are correct. If we believe that our eyes and ears give us reliable intelligence, that our memory (properly used) is generally accurate, etc., that's through no virtue or fault of our own.

3. Talk of obligations and permissions with respect to such beliefs is nonsense. First allow me to state and elaborate the thesis here: talk of duties, obligations, and even permissions, etc., to act a certain way can be made sense of only on the assumption that one is responsible for the action.[211] Indeed, it is responsibility, or potential responsibility, for the action (or the action's consequences) that makes talk of obligations, etc., coherent at all. Where one cannot speak of responsibility for an act, one also cannot with sense use normative descriptions of the act.

The best we can do is deny that `ought' or `may' apply to the act, while being quick to insist that such denial does not mean that the act is forbidden or permitted. So, e.g., if I say, "It is not the case that you are permitted to believe that p," I should not be taken to imply, "You are forbidden from believing that p." For purposes of clarity, however, I prefer to say that, with regard to acts for which we are not responsible, it does not make any sense to say we `ought to' or `may' perform them.

This thesis can be made plausible with the help of some examples.

Consider an example drawn from ethical theory. Suppose we are told that Martha is babysitting a little boy who is wading in a shallow pond – but the boy drowned when he slipped and fell, while Martha simply sat by and watched. Surely, we say, she ought to have helped. But then we are told that Martha desperately wanted to help and did her best, but she was confined to a wheelchair and physically could not get to the boy. Now it is obvious that Martha is not responsible for failing to help the boy, and accordingly it is not the case that she ought to have helped him.

Consider another example: my belief that I have two hands. I am not responsible for having this belief. Of course, it is not the case that I ought not to have this belief, because `ought' implies `can': if I cannot (under any circumstances) change my mind, it is not the case that I ought to do so. But it is not even the case that I ought to have it: the claim that I ought to believe I have hands implies that I have control of any sort over having the belief. It is in virtue of my exercising control that I can be said to fulfill an obligation.

But couldn't I say that I am, anyway, permitted to have it? No, because that claim too presupposes control over having the belief.[212] This will require more work to make plausible, but I think it can be made plausible.

"What's wrong with saying I am permitted to believe I have two hands?" you might ask. "After all, I am not violating any moral laws – to take one example of what `permission' might mean." Well, that's correct, you're not violating moral laws. But it is also important to realize that you are not responsible for this particular instance of your failure to violate moral laws. The fact that some act of yours does not violate moral laws does not by itself mean that it makes any sense to say that your act is permitted.

Different examples may help to make this plausible: a man who goes totally insane and, without understanding at all what he is doing, slaughters a busload of people is not properly said to be permitted to slaughter the people, nor is he obligated not to do so. Similarly, if you wake up with an embarrassing twitching leg, it is nonsense to say that you're permitted to twitch your leg.

It might also be helpful to consider what could be considered a platitude about the meaning of `permission': if I say that you are permitted to do (or to refrain from doing) something, I imply that you have some choice in the matter as to whether you will do it. At best it is idle for me to say that you are permitted to do something that you have no choice either to do or not to do – at best idle, I say, but in fact I think this is simply nonsense. The assertion of permission, whatever else ethicists might opine about its function, is essentially a claim about how I may exercise my freedom of choice. Correspondingly, to affirm or to deny that a permission to act exists is a claim about whether I may or may not choose to act. If it is not a matter of choice, the question of permission cannot arise.

To return now to the main argument. – From this it follows that talk of permission and duties with regard to an irresistible belief, such as that I have two hands, is nonsense. It is also nonsense to claim, for example, that one ought to believe that sense-perception is reliable, or even that one is permitted to believe this. Such talk presupposes that the belief is the sort of thing that one might be permitted or obligated to have; but one can be thus permitted or obligated only if (and because) one is indeed normatively responsible for having the belief, which is patently not the case at all. One might with just as little sense say that one is morally permitted to breathe, wake up in the morning, or perform any other involuntary bodily functions.

The same can be said for virtually all of the beliefs described as `rational' by my minimal account of rationality and for the belief that the doxastic practices that comprise Reason are reliable.

4. Conclusion: Belief in the reliability of some doxastic practices is nonjustified . Justification is, as has been commonly observed in recent years,[213] a deontological concept, i.e., one that concerns duties, obligations, and permissions. To be justified in a belief is (at least) to be epistemically permitted to hold it, or not to have flouted any epistemic duties in holding it; on a variant that is probably too strong, a justified belief is one we ought (from an epistemic point of view) to have. The permissions, duties, or obligations involved here are epistemic rather than moral; they are, nonetheless, normative properties of (mental) acts.

It follows that it makes little sense, and indeed is a category mistake, to say that we could be justified in believing that the most obviously correct uses of our basic doxastic practices – are reliable – that, e.g., the most obviously correct uses of perception do result in true beliefs. This is not to say that we are unjustified in this belief, for that would equally be a category mistake; the category `justification' does not apply to irresistible beliefs.

An analogous category error would be the insistence that a death caused wholly accidentally is either right or wrong (to any degree at all). "Lightning is wrong to strike people dead" contains a similar category error. `Wrong' is not the sort of concept that applies to natural events such as lightning.

Someone (who fails to understand the point) might object that, if there are no particular obligations prescribed for a certain belief, then one is perforce permitted to have it, and it is hence justified. That much could be correct at least in cases that immediately spring to mind. But it is not relevant to the claim I am making now, which is that if there are no particular obligations or permissions (or duties, etc.) with regard to a certain belief (because one is not responsible at all for having the belief), then it is a category mistake to say the belief either is or is not justified. Consider the lightning example: while of course it is in some sense wrong (false, if not out-and-out nonsense) to say that lightning has any "obligations" to strike or not to strike, it does not follow from that that it is true (or even meaningful) to say lightning is "permitted" to strike.

Admittedly, however, my conclusion, whatever the merits of the argument supporting it, seems extremely counterintuitive. For, if the conclusion applies to certain second-order beliefs about the reliability of doxastic practices, then perforce it would apply to beliefs that should be considered justified if any beliefs are at all. For example, on my account, it would appear that G.E. Moore's belief that he has a hand turned out not to be justified at all; nor was it unjustified. Ditto countless other irresistible beliefs that on most accounts would be considered not merely justified but absolutely certain.

Some points must be stressed forcefully if I am to remove at least some of the counterintuitiveness of the conclusion. In saying that (at least some) irresistible beliefs are neither justified nor unjustified, I am not saying that their justificatory status is neutral or evenly weighted between positive and negative; rather, I am saying that it is a mistake to say that such beliefs have any justificatory status, properly speaking, at all. That is because justification is a deontological notion and hence does not apply to (at least some) irresistible beliefs.

Nor am I saying that these irresistible beliefs possess no epistemic status whatsoever, which would of course be absurd. Indeed, I think (as I have explained at length) at least some such beliefs are rational to hold.

While these points of clarification may make the conclusion less counterintuitive, they do little to counteract the appearance that the claim is radical. If, as according to the analysis of `S knows that p' that was popular for the bulk of the twentieth century, a belief must be justified in order to be a candidate for knowledge, then all of the irresistible beliefs that I say are nonjustified are also not known, either. That seems absurd. So perhaps the claim is radical enough to be dismissed out of hand. Three further points should make it clear that that would be a mistake.

First, I am aware of no good reason for thinking that the epistemic concept that informs the analysis of `knowledge' should be justification, as opposed to any number of other related epistemic concepts. Plantinga [214] and Alston [215] (among others) have suggested, for good reasons, that justification should be dethroned from its central and privileged position in contemporary epistemology, to be supplemented by concepts described using other terms or phrases, such as `warrant' and `rationality'. Hence, even if it should turn out to be a mistake to say that I am justified in believing I have hands, it would not follow from that that I do not know, with certainty, that I have hands.

Second, even if I did decide to bite the bullet, I would not be the first (nonskeptical) philosopher to suggest that irresistible, fundamental beliefs of the sort described are not known. It is Wittgenstein again who in On Certainty opines, "We just do not see how very specialized the use of `I know' is."[216] Later in the text he explains (in characteristically oracular fashion):

Must I not begin to trust somewhere? That is to say: somewhere I must begin with not-doubting; and that is not, so to speak, hasty but excusable: it is part of judging. I should like to say: Moore does not know what he asserts he knows [e.g., that he has two hands], but it stands fast for him, as also for me; regarding it as absolutely solid is part of our method of doubt and enquiry. [217]

I would like to reserve the expression "I know" for the cases in which it is used in normal linguistic exchange.[218]

I think Wittgenstein's views differ from mine in key aspects, but he is willing, as I might be (after further argument), to say that `knowledge' might not properly apply to beliefs that are the most certain. I am willing at least to make a similar claim with regard to justification.

Third, another very Wittgensteinian point may be urged in defense of my conclusion. Namely, it coheres perfectly well with how `justified' is actually used in language. Except in philosophical contexts, one virtually never speaks of justification (or, for that matter, knowledge) with regard to the most obvious of beliefs, the sorts of beliefs that we all have no choice but to have. In ordinary use, the term appears to be reserved for beliefs for which one actually possesses some justification that one might recite (i.e., reasons, arguments); or, at the very least, for beliefs about which one typically has some choice to have.

To the non-philosopher it sounds bizarre to say, for example, "I am very well justified in my belief that I have hands." By contrast, consider: "My belief that I have hands is perfectly rational." That too would be unusual (it might be uttered in the context of a discussion about sanity), but (I think) it does not have the strange ring to it the claim about justification has. This is, I speculate, just because it is straightforwardly true in the (perfectly ordinary) sense that the belief is indicative of a properly functioning, rational mind – even if having it is out of my control. (That beliefs might be out of our control and rational nonetheless, it is worth mentioning, is a view I share with Richard Foley.[219])

VII. A Solution to the PMJ and a Refutation of Meta-Skepticism.

While justifiably held justification standards ultimately receive epistemic support from beliefs that are themselves not justified, this does not entail that no belief in a j-standard is epistemically justified. More briefly: the view that j-standards are ultimately supported only by nonjustified beliefs does not entail meta-skepticism. Even more briefly: (C) does not, given the work of sections II-VI above, entail (MSK). This is the conclusion I will attempt to establish in the present section.

Essential to the meta-skeptical case is the assumption expressed in Section I above as

(2) If S is justified in the (nonbasic) belief that p, then S has some other justified belief (or beliefs) q that supports the belief that p.

My strategy, expressed at the beginning of Section II above, was to propose that the supporting beliefs might be merely rational rather than justified. First, I will develop this proposal further and, second, apply it to the case of the Problem of Meta-Justification.

Premise (2) derives its plausibility primarily, though not only, from an unexpressed assumption: a nonbasic belief that p that is not supported by justified beliefs lacks epistemic support. But if p is supported by rational beliefs, in the sense of `rational' developed above, and these rational beliefs are nonjustified and irresistible, then p does have some epistemic support, and it is no longer obvious that p lacks, specifically, what epistemic support it would need to be justified. So, that p could have adequate epistemic support for its justification can be made plausible, as follows.

Suppose that q is offered in support of p, and q is rationally held; say q is one of the paradigm-case rational beliefs such as Moore's belief that he has hands. Rationality in the sense in question is truth-linked due to the Principle of Rationality. So, since q is rational, it is probably (and, depending on how the theory of rationality might be further developed, we could say very probably) true. But q is nonjustified, we will say, because it is a paradigm-case irresistible belief. So q's lack of justification is explained not by its having a negative epistemic status but by the fact that it is wrong, nonsense, etc., to claim that there are any obligations or permissions associated with it. In that case, it seems q is a very plausible candidate for a belief that provides justificatory support for p, despite q's not being justified itself.

This is, admittedly, not a proof that q could provide justificatory support for p. I can appeal to the plausibility of this line of argument in ordinary cases, given what I have argued in this chapter. Even if the most certain, paradigmatically rational of our beliefs are nonjustified, because they are irresistible, surely (it appears) they can nonetheless constitute adequate supporting evidence for nonbasic beliefs.

But consider what makes the obvious sort of case plausible: rationality, on account of the PoR, is a truth-linked property of beliefs. These are the most clearly rational of beliefs, and to say that a premise is rationally held is to imply that it is probably true. It is that sort of guarantee that is minimally needed for the premise to be able to give justificatory support to the conclusion.

Moreover, I wonder what sort of case a meta-skeptic could possibly make for insisting that nonbasic justified beliefs must be supported by other justified beliefs, after another compelling possibility has been spelled out clearly. On this point – though I do not claim this, either, to be an argument – I cannot, on behalf of the meta-skeptic, think of so much as a straw man to knock down.

Now let's apply these insights to the Problem of Meta-Justification the problem of how, in general, to justify standards of justification that, after all, appear to be employed in their own justification. As we saw in our in-depth exploration of the meta-regress argument, the core of the problem rests in the fact that standards license the premises (and inference step) of arguments for standards; such arguments are epistemically circular. The difficulty is that this circularity appears unavoidable. So it appears we must take some standards for granted, and in doing so, it also appears we are giving up on the PMJ and embracing meta-skepticism.

But suppose that what licenses our most fundamental arguments for j-standards are standards of rationality, and that the premises, while not justified, are rationally held. It is these standards that we take for granted; or, perhaps, we give very brief, uncontroversial arguments for them on the basis of assumptions that associated basic doxastic practices – are reliable.[220] I contend that this suggestion offers the best promise for giving a solution, of sorts, to the PMJ and a reply to meta-skepticism. I don't propose to defend any particular j-standard; rather, I wish to make a few relatively modest points (in addition to the above) that explain why, given the work of the present chapter, one can pursue the project of meta-justification unmolested by fears of circularity.

An argument is epistemically circular (as we learned in Chapter 1) when the supposition that one is justified in believing one of the premises, or that the premises support the conclusion, implies that the conclusion is true. In paradigm-case circular arguments, it is the conclusion that best explains why one is justified in believing one of the premises (or the inference step):[221] as Van Cleve's account had it, it is as though one must already know the conclusion, in order to gain knowledge of it using this argument.

What is needed, then, is an argument such that the justification (or, as we will say, rationality ) of the belief in its premises is best explained by some standard that is more fundamental than the standard (or claim about reliability) expressed in the conclusion. That's precisely what the work of the present chapter places at our disposal: while j-standards do not explain the justification of the premises of our arguments for the most fundamental of j-standards, r-standards will do the job. So we also say that belief in the premises is rational, but nonjustified, and the rationality of that belief is explained by r-standards. Properly basic r-standards, those described above (in admittedly general terms), are assumed to specify the conditions under which our beliefs are probably true; hence the rationality of belief in the premises is (by assumption) a truth-linked quality and therefore, it appears, qualified for the job of supporting (more derivative[222]) j-standards.

Here is an obvious objection to this approach. It appears that all I propose to do is to justify the most fundamental of j-standards with even more fundamental r-standards. Isn't there a problem about showing r-standards to be rationally held, though, and hence an equivalent problem of epistemic circularity that will plague demonstrations of the rationality of belief in r-standards? Surely I shall not want to propose next, for example, standards of warrant to support the r-standards, for fear of being accused of meta-meta-regressism.

There are two relevant differences between j-standards and r-standards, which are also two reasons why r-standards can be the epistemic starting-points that j-standards are not. First, r-standards specify the conditions under which beliefs are rational, and beliefs can be rational despite being irresistible and hence not candidates for justification. Second, the combination of my doxastic practice theory of rationality and the PoR has it that some beliefs are probably true simply because they are the results of properly basic doxastic practices. This is not, as I have urged at length, something for which I am under any dialectical obligation to argue. So premises rationally held can, in some cases anyway, be rational without supporting reasons; whereas premises that are justifiably believed must have some manner of justifiers.

These two points have as consequences that there is a body of irresistible, nonjustified, rational beliefs that are assumed, without argument, to be true, and correspondingly a body of r-standards such that the variety of rationality mentioned in their consequents is assumed to be truth-linked. In a very loose sense, all this prodigious assuming allows one to combine the particularist and methodist approaches to the PMJ, and to bootstrap one's way to more precisified and specialized r-standards and j-standards.

In this connection it is important to distinguish, at least in principle, between the act of justifying an r-standard – an r-standard that has already been specified – and the process of formulating a plausible candidate for justification. We have already introduced this issue in Section IV above.

Experience in the field of epistemology shows that much hard work must be devoted to the task of getting quite clear on exactly how a j-standard of some type (e.g., foundationalist or reliabilist ) should be formulated. There is no reason to think it would be any different for precisely-formulated r-standards.

But this raises a problem. Once a precisely-formulated r-standard is arrived at, we then might (if we are confused) find ourselves in the curious position of wanting to say that it is this r-standard, rather than another, that is properly basic, while the very claim that it is properly basic militates against our offering an argument for the standard. Nonetheless, claims advanced in the process of arriving at the standard can be arranged as an argument to the best explanation; we have a body of data (assumptions about rational belief) that our theory, an r-standard, is intended to explain better than other theories. But in that case, our proposed, precisely-stated r-standard does receive some support and the argument in its favor would seem to be licensed by some sort of standard governing the rationality (or justification) of beliefs formed via inductive or abductive inference.

So an r-standard – which we want to claim is a very precisely formulated, reasonably strong version of an r-standard governing, for example, sense-perception – would not be basic after all, regardless of what we would like to claim for it. I believe this problem forces us to regard the above-described "body of irresistible, nonjustified, rational beliefs that are assumed, without argument, to be true, and correspondingly a body of r-standards" as being, at best, very modest and hedged beliefs, and at worst, imprecise, rough-and-ready beliefs. For the latter sort of belief, the PoR's assumption on its behalf would not be that it is certainly correct but just that it is probably true. In any event, in the actual processes both of formulating and arguing for any very strong or precise r-standard, we will not, initially, be taking any very strong or precise beliefs for granted.

It is gratifying to note that this claim coheres with experience of twentieth-century epistemologists. When we have engaged in formulating and arguing for precisified versions of j-standards, we have often taken many prosy, imprecise claims for granted. The same would apply for a process of arriving at precisified versions of r-standards. In such a process, no doubt, too, we will also be taking a properly basic r-standard for granted – namely, one about the rationality of beliefs formed based on arguments to the best explanation.

Thus far the discussion in this section has been merely suggestive and programmatic. One might well expect, however, at least an example of how we might derive a sample j-standard using the resources of the above-described theory – short of writing another dissertation and working out all the issues involved in these suggestions. Such an example would be particularly helpful to the sort of anti-foundationalist who relies on the Epistemic Ascent Argument.[223] The anti-foundationalist would surely be reasonable to expect an example, even in a programmatic discussion of this sort. This would at least make clear what issues would need to be dealt with more fully in a non-programmatic discussion; some of the issues brought up in discussion of a sample argument for a j-standard might turn out to be extremely important.

Perhaps the most important premise in any sample argument will be a bridge principle. The principle's consequent would contain `justification' as a predicate, or an entire justification standard, while its antecedent would either not include `justification' at all or would refer only to instances of justification.

Exactly how this principle will be formulated will depend on the argument's conclusion – or, conversely, we might wish to say that what the conclusion will be will depend on what bridge principle we wish to employ. But another constraint on the principle's formulation is quite independent of what the conclusion is, viz., it must be such that its rationality would not have to be established by means of argument. For, if the bridge principle's rationality had to be established by means of argument, the argument would need some further bridge principle to arrive at the first bridge principle, and so on; in this way, a Meta-Epistemic Ascent Argument would rear its ugly head.

But nothing in the theory of rationality offered in this chapter decides the issue as to when, if ever, a belief must be established by means of argument in order to be rational, and this issue will have to be left undecided here. Nonetheless, again, in order to supply a sample argument, a bridge principle will also have to be supplied; for the coherence of the example it is simply to be hoped that the bridge principle we select for our sample argument will wear its rationality on its sleeve, as it were. This might seem to be a difficult problem simply because it might appear that there is very little that one can say about justification that fails to require argument in order to be rationally believed.[224]

There is a way to formulate a bridge principle without saying anything terribly controversial about justification, however: simply make it an instantiation of an uncontroversial inductive rule. So we might use something like this:

If for a wide variety of (given) persons who have a wide variety of (given) beliefs, (1) all such beliefs occur in circumstances c, (2) every such belief is a justified belief, and (3) no known belief that occurs in circumstances c is unjustified, then a j-standard, to the effect that if S believes that p in circumstances c, then S is justified in believing that p, has some weak presumption in its favor.

Given some such bridge principle, the sample argument would look something like this:

(1a) S1 has the belief that p1 in circumstances c.

(1b) S1 is justified in believing that p1.

(2a) S2 has the belief that p2 in circumstances c.

(2b) S2 is justified in believing that p2.

(3a) Sn has the belief that pn in circumstances c.

(3b) Sn is justified in believing that pn.

(4) No known belief that occurs in circumstances c is unjustified.

(5) If for a wide variety of persons who have a wide variety of beliefs, all such beliefs occur in circumstances c, every such belief is clearly a justified belief, and no known belief that occurs in circumstances c is unjustified, then the j-standard to the effect that if S believes that p in circumstances c, then S is justified in believing that p, has some weak presumption in its favor.

(6) Therefore, the j-standard to the effect that if S believes that p in circumstances c, then S is justified in believing that p, has some weak presumption in its favor.

One might use (6) in a further (inductive ) argument that would have as its conclusion, if S believes that p in circumstances c, then S is justified in believing that p. Alternatively, one might simply substitute the latter conclusion for (6), and claim on its behalf that, given the argument, there is some weak presumption in its favor; in that case, though, the bridge principle (5) would have to be suitably modified:

(5') If for a wide variety of persons who have a wide variety of beliefs, all such beliefs occur in circumstances c, every such belief is clearly a justified belief, and no known belief that occurs in circumstances c is unjustified, then: if S believes that p in circumstances c, then S is justified in believing that p.

To procure a fully-operational sample, obviously, some plausible "circumstances c" will have to be proposed. Here I might suggest some suitably modified version of (BSP) from Section III above. The trouble with that, however, is that a belief formed in such circumstances is rather likely to be irresistible and hence not a candidate for justification; but just for purposes of providing an example, perhaps this defect can be overlooked.

This example can help us to formulate a number of objections that a fully worked-out meta-epistemology would face. I will offer some provisional replies.

First objection. If you claim that the j-standard in (6) is justifiably believed by the epistemologist offering the argument, and it is the beliefs in each of (1)-(5) that supply the justification for (6), then we have the right to ask: "What is it in virtue of which (1)-(5) are justifiably believed?" If you supply a further j-standard, then of course, by your own lights, you will need an argument for it.

Reply. It is to be hoped that it will be plausible to assert, of the premises of a fully fleshed-out argument, that they are believed rationally according to the theory of rationality presented earlier in this chapter. Hence, what is needed is a licensing r-standard, not a j-standard.

Second objection. But unless the beliefs in each of the premises (1)-(5) are held justifiably, then, according to all accounts hitherto of how to get justified beliefs from beliefs about the premises of arguments, one cannot get a justified belief in (6) from beliefs in those premises.

Reply. This objection correctly identifies one requirement of a more fully-developed account: we must restate the conditions under which we can get a justified belief from an argument in such a way that justified beliefs in the premises are not necessary and in such a way that rational beliefs will provide adequate support.

An applicable point was insisted upon above, at length: there is no good reason to think that a belief that supplies justificatory support needs to be justified in order to supply such support. No doubt any supporting belief must have positive epistemic status of some sort that is truth-linked. Moreover, clearly, it remains an important but as yet unanswered question exactly what conditions they will be under which some variety of positive epistemic status, such as rationality in our sense, will provide the needed support. But by my lights it is sensible enough to think that such an account is forthcoming, because, after all, among the irresistible rational beliefs are some that are at least as certain as any justified belief.

The latter defense can be further bolstered by the following consideration. For purposes of supplying justificatory support, on deontological accounts of justification, what merely rational beliefs lack that justified beliefs have is a sort of built-in deontological component. That is, one can say that a belief justifiably held is (for instance) permissibly held, while one cannot say this of a merely rational belief. Admittedly, for purposes of supplying justificatory support, perhaps there is something more that justification supplies that rationality does not. But if not, then the only reason justification could be thought to be required, in order to give adequate justificatory support via argument, is that deontological component. And in that case, why insist that rational beliefs cannot supply the needed support? Why couldn't it be the case that what "permits" us to have justified beliefs are rational beliefs, about which it makes no sense to say that we are "permitted" to have them? Why should it make any difference whether the supporting beliefs are themselves permitted?

Third objection. The argument works only if each of the premises, including claims that various beliefs are justifiably believed, are rationally held according to the sense we have advanced. But how plausible is it to say that claims about when beliefs are justified are rationally held?

Reply. This is admittedly a difficult objection, but we can urge one point in reply: in the field of epistemology, we all do, after all, depend on unsupported "intuitions" about when beliefs are justified or not. For arguments for our initial, weak j-standards, we will want to employ only the most obviously justified beliefs. It is not implausible to suppose that one could be rational, in our sense, in believing that those beliefs are justified.

So much for the process of arriving at at our initial j-standards with the help of r-standards. Again, the above is only a sketch of how the argument might go.

I propose, in addition, that derivative, specialized, narrowly-focused r-standards and j-standards could be, and in a loose sense have been, developed and justified through a method of wide reflective equilibrium – that begins with properly basic r-standards.[225]

The above-mentioned body of prosy, imprecise beliefs (against which initial r-standards are to be tested) again plays a prominent role. But how the more specialized standards are tested varies considerably depending on the field of inquiry. By way of broad generalization, one may say that new refinements to modest, generally-applicable standards are proposed, and those refinements are tested according to consistency with other established standards and rational background beliefs.

A helpful, realistic description of such a refinement procedure would involve researching the actual historical development of scientific, professional, and academic methods of various sorts, via a thorough study of intellectual history. There's little point to speculation about how the arrival at modern standards is to be reconstructed, and thence (presumably) justified, when many relevant facts are to be learned in a study of intellectual history, e.g., the history of science.

For example, if one wished to reconstruct the path – no doubt a path that resembles the give-and-take Method of Reflective Equilibrium – whereby we arrived at the exceedingly fastidious standards according to which contemporary philosophers interpret the history of philosophy, the place to begin would be via some very simple, prosy standards about testimony, as well as some texts in which historians of philosophy began taking each other to task for their clumsy misinterpretations of dead philosophers. The simple, prosy standards about testimony (e.g., a principle of charity: "Interpret someone so that he says something true, if possible") informed the development of modern standards. When it became clear (particularly in the early twentieth century) that certain easy (and misleading) formulations of historical positions would not wash, new standards had essentially been put in place: among other things, it became necessary to consult the texts themselves. The more enlightened historical interpreters could always point out to their more negligent peers that if we have no historical text in which so-and-so actually held that p, then the claim that so-and-so held that p requires some argument. Such a change in attitude could be described as a step in the Method of Reflective Equilibrium.

Similarly, a study of the history of logic makes fairly plain that the path that led to (the various) contemporary standards of deductive inference is one in which various refinements were made to relatively crude, but intially quite plausible rules, initially stated by Aristotle. Along the way, refinements were tested according to the simplest and least controversial sorts of principles: if one can find an instance of an argument form where the premises are all true and the conclusion is false, the argument form is not valid.[226]

One might object to all this that we are interested not in how, historically, we arrived at the j-standards that we now employ but in how we might show that those standards are in fact justified, or how exactly those standards should now be formulated. Again, if I wish to say that I have solved the problem of induction (as well as a bunch of similarly-structured epistemic problems), it will not do simply to say that some standard or other of induction describes rational belief-formation. I should produce the exact standard or standards.

But the above discussion is intended as a response to this very objection: in order to arrive at a formulation of "the exact standard" for any field of inquiry, surely an in-depth study of actual standards and probably actual procedures of refinement will be required. In any case, it is not encumbent upon me, in the present discussion, to produce such a study in order to defend my line of argument: I think it is adequate to point the way to a solution and then as much as needs to be done, as far as describing refinement procedures for j-standards, has been done. If this means I have failed to defend a complete solution to the PMJ, then so be it. What I have done, I maintain, is to show the way to a solution.

If a solution to the PMJ is indeed permitted by this chapter's work, it is clear that an answer to meta-skepticism is in the offing. The relevant question is whether the following claim entails meta-skepticism:

(C) For any given justification standard, J, acceptance of it is either not justified, or ultimately receives support only from beliefs that are themselves not justified (regardless of how many intervening beliefs there might be between the ultimately supporting beliefs and the standard).

If the process of supporting j-standards begins with r-standards and basic beliefs that are nonjustified, the right disjunct of this claim is satisfied: j-standards ultimately receive support from "beliefs that are themselves not justified." But such support can be sufficient for us to be able to say that those standards are justifiably believed. Hence (C) does not entail meta-skepticism.

VIII. Conclusion.

To gain some perspective, it will be interesting to compare and contrast the solution to the PMJ developed above to what Alston says in the last chapter of The Reliability of Sense Perception about what is essentially the same problem. Some of Alston's conclusions are very similar to my own, but the differences are worth noting.

Alston comments on a position similar to the one advanced in this chapter:

[A "naturalized" epistemologist] might hold that the reliability of our familiar basic doxastic practices is just a rock-bottom commitment from which there is no appeal. It is impossible to find anything more basic on the basis of which this commitment could be evaluated. I find this claim quite appealing, and it will play a major role in the response I shall shortly be advocating, although it is set there in a larger context that involves a kind of justification of it, as well as suggestions of how this commitment can be tested to a certain extent. But a totally uncritical acceptance of our customary practices, without any provision for rational rejection or modification, I find quite indefensible, provided, as I shall be arguing shortly, there is a possibility of rational criticism.[227]

Of course, in the present chapter I have not argued for a "totally uncritical acceptance of our customary practices," but for the acceptance of our most basic practices, those without which further investigation of other practices would be impossible. More derivative doxastic practices – specialized practices used in the various professions, sciences, etc. – should not, of course, be accepted uncritically (in some circumstances, anyway). There is no particular reason (considered in this dissertation, anyway) for being very conservative about any established derivative doxastic practices. If they are subject to noncircular confirmation or disconfirmation, then from an epistemologist's point of view, the fact that they are in common use might only bespeak some presumption in their favor.

So while Alston's remarks do not address the precise position I have developed, in fact I happily find myself in agreement with his general approach, one of "taking it to be rational and proper to engage in our customary doxastic practices [or a small subset of them, anyway] without having, or even being able to have, any positive noncircular reasons for supposing them to be reliable."[228] Where we differ is on the question of what makes (what I call) basic doxastic practices rational. My claim, following the likes of Reid and Strawson, is that they constitute rationality itself. Alston has a fine discussion of Reid's views[229] but does not adopt a specifically Reidian view of rationality.

Alston says that Reid takes "all of our established doxastic practices to be acceptable as such, as innocent until proven guilty,"[230] but Alston qualifies the term "established" as meaning "firmly rooted in our lives, practices which we could abandon or replace only with extreme difficulty if at all."[231] Established doxastic practices are hence, he says, "practically rational." Alston does not actually supply us with a separate account of `practical rationality'. The context makes it clear that he does not simply identify practical rationality with established doxastic practices: after all, he conceives of his Section ii in his final chapter as providing us with "A Practical Argument for the Rationality of SP [Sense Perceptual Practice]."

Then Alston examines the relationship between practical rationality and reliability.[232] Alston claims that the practical rationality of a practice does not constitute evidence for its reliability, but rather, "I believe that in showing it to be rational to engage in SP, I have thereby, not shown SP to be reliable, but shown it to be rational to suppose SP to be reliable."[233] But Alston makes it clear that the sense of `rational' in which he thinks he has shown it to be rational to believe that SP is reliable is, again, practical rationality – which, again, seems to amount to saying that such a belief is "firmly rooted in our lives" and difficult to change.

By contrast, my own view on the rationality of assuming SP and other basic doxastic practices to be reliable is expressed plainly in my discussion, above, of the PoR. Since I say I assume that properly basic doxastic practices, which constitute minimal rationality, are reliable, I would be uncomfortable if one were to say that I argue, "SP is rational; therefore, it is reliable." If anything the argument goes in the opposite direction.

But what strikes me as most plausible is that, in point of conceptual and semantic fact, `rationality' and `Reason' in one sense denote a set of basic doxastic practices; and it so happens that we cannot resist believing that those practices are reliable, which belief is expressed in the PoR. This is one, but only one, reason I gave for thinking that it is defensible, within the context of work in epistemology, to advance something like the PoR without argument. This is not to claim that belief in the principle is rational. Surely it is rational, but to defend that view that I think I would need an account of `rationality' that would apply to such matters. The account I have supplied does not, I think, apply to the principle: acceptance of the principle surely does not issue from a properly basic doxastic practice.

Whatever the differences, the similarities between my proposed solution to the PMJ and Alston's are striking.[234] Still, if the work of this chapter lacks originality to some degree it is not because it borrows from Alston, but because it borrows much from Reid. Indeed, I regard this chapter (and several individual points from Chapters 1-3) as applying and developing some of the fundamental doctrines of Reidian epistemology to a modern formulation of a very old problem. I am hoping that this will be welcome, anyway. Many contemporary epistemologists – if not other specialists – have fortunately unlearned that old, uninformed prejudice, that Reid's philosophy of common sense is unsophisticated and not worth study.

I have not explored the consequences of many points that I have been brought out, because doing so would have taken us far too far afield from the topic of how one might accept (C) without embracing meta-skepticism. I invite the reader to consider again the contentions advanced in this chapter:

For any given justification standard, J, acceptance of it is either not justified, or ultimately receives support only from beliefs that are themselves not justified (regardless of how many intervening beliefs there might be between the ultimately supporting beliefs and the standard).

There is a useful sense of `rationality' according to which certain basic doxastic practices constitute rationality; beliefs might be rational because they result from such practices.

Such doxastic practices – are `basic' in the sense that their reliability cannot be defended using any combination of any other practices, and `properly' basic in the sense that they result in beliefs that are as obviously true as any that we possess.

This theory of rationality essentially endorses and expands Strawson's solution to the problem of induction.

According to the Principle of Rationality, beliefs formed in accordance with those practices that constitute minimal rationality are probably true.

The latter principle may be advanced without argument – which is a move that deserves to be taken seriously.

There is a category of belief to which it is a category mistake to apply the terms `justified' and `unjustified'; they are nonjustified beliefs. Such beliefs are irresistible, and hence not candidates for the permissions and obligations that are part and parcel of the concept of justification. They can, however, be rational and indeed absolutely certain.

Justification standards may be argued for using noncircular arguments that are licensed by standards of rationality, on the basis of beliefs that are rational but nonjustified; hence the results of Chapters 1-3 need not be construed as supporting meta-skepticism.

Less basic practices, including the highly specialized doxastic practices that constitute the modern, complex, scientific conception of rationality may be defended on the basis of the basic ones, in a bootstrapping fashion similar to that advocated by Max Black and the meta-coherentists.

Taken together, these conclusions have a variety of important consequences worth investigating; but I could not reasonably claim to have firmly established many of these conclusions. Still, at least they appear to constitute a coherent, broad, powerful meta-epistemological theory that might stand further, long-term exploration.

Next Section