CHAPTER 3

OTHER ATTEMPTED SOLUTIONS TO THE PROBLEM OF META-JUSTIFICATION

I. Introduction: Coherentism and Meta-Coherentism.

Meta-coherentism is the view that mutually supported justification standards exist. On the one hand, like self-supportists, meta-coherentists believe that, ultimately, standards do give support to themselves. But, on the other hand, meta-coherentists also believe that standards do not support themselves via single, relatively simple arguments like track record arguments, but via an entire doxastic system containing a web of arguments – arguments themselves licensed by a variety of other standards. So meta-coherentists react to the Problem of Meta-Justification differently, but not entirely differently, from self-supportists.

Ernest Sosa conveys very well the attraction that meta-coherentism has over self-supportism:

How else indeed could one determine the reliability of one's sources of belief than by considering the accuracy of their deliverances and assessing them on that basis?...

What is wrong in the newspaper case [in which we accept new reports because they cohere with old reports, which we have taken at face value in the past; an example of self-support ], even as a case of simple reasoning is, it now appears, the narrowness of one's purview in judging the newspaper reliable simply on the basis of a set of data one knows to be remediably and relevantly too narrow; namely, the reports of that very newspaper accepted at face value.... So the circle can perhaps be widened to make it after all benign. Perhaps comprehensive coherence is after all a legitimate court of appeal.[99]

Meta-coherentism has considerable intuitive appeal when one considers simply that we are restricted in all our inquiries to using whatever background assumptions and cognitive tools we have at our disposal. It would seem then that any justification of those assumptions and tools will require the use of those same assumptions and tools. Hence the following has some intuitive appeal: the circularity involved in presenting a coherent system, in an attempt to solve the PMJ, should be permitted if for no other reason than that it is unavoidable. As I will argue, this intuitive appeal is deceptive, and meta-coherentism must be rejected.

On any pure coherence theory of justification, what justifies us in accepting any belief is simply the belief's coherence with, and place within, a coherent system of belief. "Coherence" means at a minimum consistency, but usually also describes such relations as confirmation and explanatory simplicity.[100]

Coherentism thus characterized is clearly not equivalent to meta-coherentism. Meta-coherentism does not, at any rate, entail coherentism. One might be a meta-coherentist, thus holding that mutual supports relations exist among j-standards, and yet be a foundationalist on the object level. In that case, one simply holds that the reason that one's foundationalist standards are justifiably believed is that they cohere with each other via their various outputs and via background beliefs. But one would of course have to make specific (and perhaps dangerously ad hoc) exceptions for the justification of foundationalist standards within a more general foundationalist theory of justification. Beliefs in standards are, after all, candidates for justification just like any other beliefs, and so if they are singled out as being justified in a way that is different from other beliefs, that special treatment would have to be well-motivated.

On the other hand, it is more natural, if one is a coherentist, also to accept meta-coherentism. On most versions of coherentism, a belief's justification is due at least to the fact that the belief is contained in a coherent doxastic system; and usually a system is considered coherent only if there are mutual confirmations among the beliefs. It would follow that on most versions of coherentism, at least, any given j-standard is going to add at least some degree of confirmation, and thus warrant, to other standards, and receive warrant in return.[101]

However all that might be, to the extent to which meta-coherentism resembles, or entails aspects of, coherentism, the several well-known criticisms of the latter may be applied to the former.[102] Alston goes so far as to rule meta-coherentism immediately out of court on the strength of the criticisms against coherentism:

Coherence theory holds that individual beliefs gain positive epistemic status... by virtue of being involved in a total system of beliefs that is coherent.... Reciprocal support is the rule rather than the exception. From this standpoint there is nothing disturbing about the circle involved in using perceptual beliefs to support the principle that sense perception is reliable.... However, for purposes of this essay I am setting aside coherence theories without a hearing. I take it that the live possibility of a multiplicity, perhaps an indefinite multiplicity, of incompatible but equally coherent systems of equal comprehensiveness is sufficient to show that internal coherence cannot be the whole story of what gives beliefs a positive epistemic status.[103]

Alston, then, rejected meta-coherentism for the same reasons that he rejected coherentism. While I share Alston's disdain for coherentism, I am also aware that desperation in the face of the Meta-Regress Argument might push us to try to meld some unobjectionable form of meta-coherentism with foundationalism.[104] In that way we might try to avoid the typical objections to coherentism – with what success it is hard to say.

II. Meta-Coherentism and Epistemic Circularity.

A further argument, specifically against meta-coherentism (in addition to those against coherentism alluded to above), would be helpful. And one is possible. To wit: meta-coherentism is committed to the benignity of epistemic circularity; but epistemic circularity is in fact vicious; therefore, meta-coherentism is false.

It is common to deny that the circularity involved in a system of mutually supported standards is vicious. Again, as Alston says, "a coherence theorist will not be at all disturbed by the pervasiveness of epistemic circularity, since circularity (of the right sort) holds no terrors for him."[105]

Alston is right: epistemic circularity does not bother some representative coherence theorists. Nelson Goodman, in advocating a deductive justification of deduction, and an inductive justification of induction, with a goal of what is now called "reflective equilibrium," writes, famously, that the resulting circles are not vicious but "virtuous."[106] Rescher, in one of his many books advocating a coherence method of theory confirmation, wrote, "There is thus no reason to concede that the circle at issue is vicious or otherwise vitiating."[107]

But again, the PMJ is likely to push some to desperate measures – including the denial that meta-coherentist justifications involve epistemic circularity at all. Meta-coherentists are hard pressed to deny, plausibly, that the existence of mutually supported standards does entail the existence of viable epistemically circular arguments.[108] So we should, to make the full force of the problem felt, close off all possible avenues of escape. How, then, can it be shown that meta-coherentist justification does indeed make use of epistemically circular arguments?

As follows. According to meta-coherentism, mutually supported j-standards exist. In other words, meta-coherentism has it that, when S is justified in believing some standard, J1, S has an argument A1 for J1; but A1 is licensed by another standard, which itself is justified by another argument; and we iterate the chain, or web, of argumentation as needed; then we conclude with an argument A2 for a standard J2, where J2 licenses the most recent member of the chain of argument that leads to J1; but A2 is itself licensed by J1. That is the logical structure of meta-coherentist justification.

Now, it is far from clear that, according to the definition of `epistemic circularity' given in Chapter 1, A1 by itself is epistemically circular. That would, after all, obviate the need of supporting its premises with other arguments. But an effective indirect argument, for an equally devastating claim, is possible. Let Ai be the argument constructed out of the series A2,..., A1, with the ultimate conclusion J1, and supplemented as follows. Ai is divided into some number of intermediate arguments, the conclusion of each being used to argue explicitly for some premise or the inference step of the next. If indeed A1 is justifiable on the basis of the arguments A2,..., A1, then I claim (1) such an Ai is constructible, and (2) Ai could justify J1.[109] But in this case J1 is self-supporting; hence, we may use an argument used in Chapter 1, Section VII, to conclude that Ai is epistemically circular. The upshot is that the chain of meta-justificatory arguments for any mutually justified standard can be strung together so as to make the standard self-supporting, whereupon the epistemic circularity involved in the meta-justification becomes more evident.

Hence, if a standard is mutually supported, then that same standard can be self-supporting. But the arguments of Chapter 2, if sound, show that no standard can be self-supporting. So by modus tollens no standard can be mutually supported. Therefore meta-coherentism is false. By this accounting, all of the arguments against self-supportism apply equally well to meta-coherentism. This is more than adequate grounds on which to reject meta-coherentism, I think.

But meta-coherentists are apt to bring out certain considerations, not addressed in Chapter 2, with the hope of mitigating the viciousness of circularity. These considerations do not always appear in the context of the PMJ per se, but in discussions of theories and methods that appear to be committed to meta-coherentism – especially the Method of Reflective Equilibrium. So, next we will turn to some questions surrounding that method, in order to determine whether its use might, somehow, mitigate the viciousness of epistemic circularity.[110]

III. Can MRE Help to Mitigate the Viciousness of Epistemic Circularity?

First, however, we should come to an understanding of what the Method of Reflective Equilibrium (MRE ) is that is adequate for our purposes.

The phrase `reflective equilibrium' was introduced by John Rawls [111] to indicate a mental state achieved after undergoing a process that, roughly put, involves adjusting and re-adjusting accepted moral principles and considered judgments about the morality of particular actions so that they cohere with each other. Once this state of coherence has been achieved for a person, that person is said to be in `reflective equilibrium', and in particular, `narrow reflective equilibrium'. `Wide reflective equilibrium' means a state achieved after moral principles and judgments are made to cohere not only with each other but also with a wide variety of background assumptions.[112]

The basic notion of the MRE, without being so named, was introduced (at least to the current generation of philosophers ) by Nelson Goodman. In a brief discussion in Fact, Fiction, and Forecast, Goodman writes:

How do we justify a deduction [a particular deductive argument]? Plainly, by showing that it conforms to the general rules of deductive inference.... Yet, of course, the rules themselves must eventually be justified.... But how is the validity of rules to be determined?... Principles of deductive inference are justified by their conformity with accepted deductive practice. Their validity depends upon accordance with the particular deductive inferences we actually make and sanction. If a rule yields unacceptable inferences, we drop it as invalid.... The point is that rules and particular inferences alike are justified by being brought into agreement with each other. A rule is amended if it yields an inference we are unwilling to accept; an inference is rejected if it violates a rule we are unwilling to amend.[113]

Assume that we have at our disposal some stock of "intuitions," or "considered judgments," as well as tentative generalizations about how some evaluative concept is to be applied. Then, to arrive at a justified generalization about the applicability of the concept, we need only – to put it roughly render this set of intuitions and generalizations as coherent as possible while retaining as many of the original members of the set as possible. Whenever we discuss generalizations about the applicability of nearly any important evaluative word, and the question of the justification of such generalizations arises, MRE, or something much like it, often naturally suggests itself. MRE has been thought applicable not just to the confirmation of moral principles, but also to deductive and inductive inference rules, and most notably for our purposes, of course, to standards of epistemic justification.

As Paul Moser writes:

Moral theorists sometimes appeal to a strategy of `reflective equilibrium' to justify an ethical theory. It seems that we can use an analogue of this strategy in epistemology. The analogue in epistemology begins with one's having certain tentative, revisable epistemic intuitions about particular instances of justification and knowledge. It then involves one's asking what explanatory epistemic principles, if any, entail those putative instances of justification and knowledge.... The epistemologist often works back and forth between epistemic principles and considered judgments. One sometimes revises considered judgments about particular cases in light of effective explanatory principles;...[114]

We can see here that Moser's account of the justification of epistemic principles closely follows Goodman's account of the justification of inference rules.

Obviously, just reading the passages cited above, many questions are left unanswered. What is a "tentative, revisable epistemic intuition," and how do I distinguish it from other beliefs? How do I decide which "explanatory epistemic principles" to test against these intuitions? What exactly is involved in the "back and forth" work? Exactly when am I warranted in dropping an intuition, or a principle – i.e., how do I know when the coherence of my doxastic system is strengthened when I do so?

Because we will not examine the merits of MRE in general, it will be unnecessary for us to answer such questions here. We need only answer two questions. (1) Does MRE, somehow, involve meta-coherentism? (2) Insofar as it does, does it offer some unique way to mitigate the viciousness of epistemic circularity? Fortunately, to address these questions we do not require any detailed account of MRE and how it might proceed.

To take up the first question, then: does MRE entail meta-coherentism? Strictly speaking, the answer is no. MRE is a method of formulating j-standards. It is a method of inquiry, not a theory about when the act of justification succeeds. MRE is intended to state a more or less systematic way of arriving at particular sorts of claims, and as such it is not, or not necessarily, an account of how to justify those claims. Moreover, as epistemologists should know, an account of justifying, or of the act of justification, is quite a different thing from an account of justified belief, or that in virtue of which a belief is justified. So MRE is in fact two steps removed from the latter sort of account. To call it, without further ado, a kind of meta-coherentism would be a mistake.[115]

In fact, that is not the mistake that some people have made; rather, some have thought, mistakenly, that MRE is a kind of, or is committed to, foundationalism.[116] Criticizing that view, Michael DePaul nicely sums up:

I see the Method of Reflective Equilibrium as being first and foremost a method. It is a heuristic device for organizing our moral beliefs, a manner of conducting our moral inquiries. Foundationalism, on the other hand, is primarily a type of account of the epistemic status of our beliefs. Hence, foundationalism and reflective equilibrium are not really positions on the same topic, although they are surely positions on related topics.[117]

Nonetheless, it is common for MRE to be regarded as entailing, or being otherwise closely connected to, coherentism. Sosa puts it this way: "Wide equilibrium seems equivalent to a pure coherentism which at any juncture would always opt for the most harmoniously and comprehensively coherent view available."[118] DePaul describes MRE as a "coherence method" and specifically argues in this article that MRE is compatible with coherence theories of justification.[119]

That's fine, but how exactly is MRE, as a method, associated with a theory of justification? Is there some way to show that MRE somehow entails that standards that result from its use are justified in accordance with meta-coherentism? Probably not; this is a difficult issue, and one we needn't fully engage here anyway, because the point of bringing up the MRE at all is to determine whether it offers any resources whereby the epistemic circularity of meta-coherentism is mitigated. So let us focus on that issue.

Sosa wrote an article on the topic of the relation between method of inquiry and theory of justification. Sosa proposed this account:

A belief is justified (warranted, reasonable,...) iff it is obtained or supported by appropriate use of an adequate organon.[120]

Indeed, MRE might be regarded as such an "organon" by some of its proponents. Now, whatever you might think of Sosa's account of justification – some externalists will object to it right away at least it does forthrightly state a view on the relation between method and justification (including, evidently, meta-justification ). I will argue now that even if Sosa's view has the relation between MRE and (meta-) justification correctly stated, MRE still will not help to mitigate the viciousness of circularity.

So let us assume that it is claimed for MRE that it results in an allegedly justified belief in a justification standard J, which belief justifies (in part) beliefs in other standards, and which is itself justified (in part) by those beliefs in the other standards. So MRE is said to result in a mutually justified belief that J, and if such a belief is (successfully) mutually justified, then meta-coherentism is true. We then charge this alleged meta-justification with epistemic circularity; how can the proponents of MRE try to counter the charge?

Let us begin with Goodman again:

This looks flagrantly circular. I have said that deductive inferences are justified by their conformity to valid general rules, and that general rules are justified by their conformity to valid inferences. But this circle is a virtuous one. The point is that rules and particular inferences alike are justified by being brought into agreement with each other.... The process of justification is the delicate one of making mutual adjustments between rules and accepted inferences; and in the agreement achieved lies the only justification needed for either.[121]

This appears to be everything that Goodman has to say about the circularity problem. As far as I can tell, all that Goodman does is to describe this peculiar sort of circularity between rules and inferences, and then claim, "in the agreement achieved lies the only justification needed for either." That claim perhaps commits him to some sort of meta-coherentism, but it certainly does not explain how such meta-coherentism alleviates the circularity problem. Despite its seeming lack of substance,[122] this passage – its view of circularity has been very influential.

Nicholas Rescher does not defend MRE per se but a method of which MRE might be an example. In Rescher's method, which he calls "instrumental justification," by his own admission there is a circularity, one that is virtually the same as the circularity involved in MRE (and thus, on our present assumption, in meta-coherentism ).[123]

Rescher claims that this sort of circularity is benign ("nonvicious") because "the argumentation is comprehensively systematic, placing its several elements into a coordinate framework which unites them within one over-all nexus of mutual substantiation." He also says that this sort of justification is not "unidirectional" but instead "a systematic union" of "method" and "thesis."[124] So far, Rescher appears to have no more insight than does Goodman about why meta-coherentism does not involve any vicious circularity.

Later Rescher does say something usefully different: "There is thus no reason to concede that the circle at issue is vicious or otherwise vitiating, for what is actually involved is simply a feedback process of a type nowadays familiar from the study of self-regulatory systems" like "a self-evaluating servomechanism."[125] Rescher explicitly affirms that the reason his method of meta-justification is not viciously circular is that it is a "cyclic process."[126] Repeating a claim in the course of carrying out a process is not necessarily viciously circular – or, indeed, circular at all.

Rescher sums up this approach as follows:

All of these structurally analogous problems [different instances of epistemic circularity ] have a structurally analogous solution – viz., that the paradox arises from taking a strictly static point of view, and that these difficulties vanish when [one] regards the issue from the dynamic perspective of a cyclic feedback process.[127]

There is surely some initial plausibility to this view. If justification is regarded as a "process," rather than something "static," then perhaps there is nothing wrong with a claim that is repeated several times throughout the process.

This will not do, however. As I explained, a method of inquiry must not be mistaken for a theory of justification. But it appears Rescher has not only mistaken one for the other, he has gone on to say that since one can repeat the same claim while engaging in a process, without any vicious circularity, therefore there is no vicious circularity involved in the assertion that the process results in a justified belief.

This does not follow. Here is a dilemma. On the one hand, following Rescher's method is alleged to generate a standard's justification. But in that case, there is a vicious circularity, if sufficient weight is given to my arguments in Chapter 2. Or, on the other hand, following the method is not alleged to constitute a justification, but only a way of generating theses, in which case there is no circularity at all, but also no attempt at a meta-justification.

Rescher's point and my answer to it can be stated differently. First Rescher's point: "Yesterday I argued that q on the basis p. Today I argued that p on the basis of q. You will see a vicious circularity here only if you view all of my arguments as part of one static system. But arguing is a process that changes from day to day. There is no circularity if we focus on one part of the process at a time, one day at a time."[128]

This line is a little like what the polygamous airline pilot says – you know, the one who claims to be monogamous. He's got only one wife at a time! He doesn't see them all at once! Less flippantly, I respond with a dilemma. If we regard the arguments in which p and q occur as part of one "static" system that is supposed to give expression to the justification for p and q, then clearly, given the conclusions of Chapter 2, there is a vicious circularity involved in the system. Nothing that we have seen from Goodman or Rescher controverts that. But if we take Rescher's advice and shift our focus to what transpires just "today," to limited parts of the process of argumentation, then we never, on any "one day" as it were, succeed in showing that standards of justification are accepted justifiedly. After all, Rescher would have it that justification is due to the coherence of an entire system.

It looks like Rescher is trying to avoid the consequences of his meta-coherentism. The meta-coherentist says that justification is due to mutual support relations between standards; but we cannot properly grasp and evaluate those relations unless we regard an entire meta-justificatory system at once, as something "static."

To sum up. The point of this section's discussion is that certain methods of inquiry (loosely) associated with meta-coherentism – the Method of Reflective Equilibrium, Rescher's "instrumental justification," among others are frequently treated as ways to mitigate the viciousness of epistemic circularity. But we are typically left in the dark as to how the use of these methods renders the circularity any less vicious. Rescher and others appear to have convinced themselves that repetition of a claim while engaging in an inquiry does not necessarily involve any circular reasoning at all. That is true enough, but what we are interested in are meta-justifications, not simple inquiries. Meta-ethicists and other proponents of MRE might, then, wish to use the method's resources to defend meta-coherentism against charges that it permits vicious circularity; but, as far as I have been able to discern, the method lacks adequate resources for the job.

IV. Does the Size of the Circle Matter?

There's another, fairly common way to defend meta-coherentism against charges of vicious circularity. Briefly, the strategy is to admit that small circles are indeed vicious, but if a circle in question is "large" enough, it is virtuous.[129] Let us introduce this strategy with an analogy Sosa discusses; the reliability of sense perception can be understood by analogy with the reliability of a newspaper.[130]

The inaugural issue of the Podunk News just arrived and I want to know if it is a decent newspaper. Perhaps I don't get out very much and until now haven't read too many newspapers. But what the News says is internally consistent from day to day; and, being an unfortunately gullible sort, I tend to believe what I read. Today I read, "There were 27 murders last year in Podunk." So I believe that; then, when I return again to the pages of the News tomorrow, I find my new belief confirmed again. (E.g., "Mayor Decries Rising Murder Rate.") After enough cases like this, I conclude that the News is reliable. That's an example of a "small" circle: I believe that p because the News says that p and the News is reliable; but I believe that the News is reliable, in part, because the News says that p (which I believe).

Similarly, the world our senses present to us is "consistent" – apparently rule-governed – and we simply tend to believe that what appears to our senses exists in about the way it appears. It visually appears to us that p, and the sense of sight seems reliable, and so we tend to believe that p. But, armed with the belief that p, upon finding further (visual) confirmation of p, we conclude that our sense of sight is, generally, reliable. That is, essentially, what is going on in a track record argument, and it runs about as small a circle as in the case of the Podunk News.

But now the meta-coherentist leaps to the defense of these naïve newsreaders and perceivers. These accounts of what's going on, he says, are oversimplified. It is not just from the News itself that I learn about the world at large. I also catch the news on TV and radio, and occasionally have an opportunity to speak to people directly involved with news events. The reports are generally consistent. This wide array of consistent data is best explained by the claim that the News is generally telling the truth. If among those data to be explained are claims made by the News, that's all right; the argument might be circular to that small extent, but the data, together with the claim that the News is reliable, are coherent. Any circularity is to be expected and surely does not impugn the justification of this claim about the reliability of the News – as long as the rest of the argument, in which the circularity is embedded, is large enough.

Similarly with sense-perception. It is not only data about present visual experience that support the thesis that visual perception is reliable; a wide array of data from other sense modalities and from other beliefs (e.g., the belief that our faculties would have evolved in such a way as to provide us with mostly true beliefs) corroborate the conclusion we might draw from raw visual data alone.[131] The argument might still be circular, but only due to a relatively small portion of the argument – i.e., that portion in which beliefs about what I see both support and are supported by the claim that visual sense perception is reliable.

The relevant question to ask is: what difference does it make how large the circle is (i.e., how large the argument is in which the circular reasoning is embedded)? Surely the assertion, that large circles are virtuous because they are large, needs support – perhaps a minimal amount of support for those grasping for ways to defend meta-coherentism, but some support nonetheless. But the literature is strangely silent on this point. At best, we get hints. Teasing out the hints will involve us in some unavoidable, but (it is to be hoped) forgivable, obscurity.

For example, Sosa writes:

How else indeed could one determine the reliability of one's sources of belief than by considering the accuracy of their deliverances and assessing them on that basis? In the absence of cognitive science, there appears no other way. … So the circle can perhaps be widened to make it after all benign. Perhaps comprehensive coherence is after all a legitimate court of last appeal.[132]

So, suggests Sosa, we have no choice but to admit epistemic circularity. This is the best we can do in evaluating the reliability of, e.g., sense-perception; so it is best to assume that any circularity involved is acceptable.

But, evidently, these "wide circle" arguments are not the only "court of last appeal" we might devise. There are, after all, the other options in the Meta-Regress Argument to consider. So if the suggestion is that we must learn to live with epistemic circularity because it is the only option open to us, then given what we have already explained, the suggestion may be rejected, because there are indeed several other options to consider (and canvassing them is precisely what this dissertation is about).

From conversations with various philosophers and from other obscure passages in various defenses of meta-coherentism, I gather that there is a second view on why the size of the circle should make any difference. Namely, the addition of the epistemically circular element – that is, of premises that are themselves justified in accordance with the conclusion – adds to the overall coherence of the argument. (Henceforth that "element" of the argument will be known as the offending premises .) More precisely, the offending premises are neatly (but unsurprisingly) explained by the conclusion. So, what complaint should we have about the addition of data that strengthens the argument? Indeed, if the set of the other premises, which do not depend on the conclusion for their justification, is large enough, then any "harm done" by the circularity is small.

This second view could stand a bit more elaboration. The idea is that an epistemically circular argument's coherence is increased by the offending premises; and this increased coherence somehow overrules, or outweighs, or gives us adequate reason to ignore, the fact that the offending premises make the argument circular. One might understandably wonder how it is that we can simply ignore a clear defect in an argument at all; but humor our meta-coherentist on this point for a bit.

About this line of argument one need only ask: if the contribution of the offending premises (i.e., those that are justified in accordance with the conclusion) is small enough that little "harm" is done by the resulting circularity, then why not remove the offending premises altogether? For example, why not try to construct an argument for the reliability of sense-perception without using any beliefs gained by the use of sense-perception?

Suppose one is tempted to answer, "Good idea – indeed, why not?" Then one contemplates abandoning meta-coherentism altogether for some other solution to the meta-regress problem listed in Chapter 1. Meta-coherentism by its very nature involves epistemic circularity, as demonstrated earlier in this chapter.

So suppose, instead, that one remains true to meta-coherentism; one opts to say that the contribution of the offending premises is essential and that there is no other way to defend the reliability of sense-perception, or any number of basic j-standards for that matter. But then, clearly, one is abandoning the position that little "harm" is done by the resulting circularity. After all, if the success of an argument for the reliability of sense-perception crucially depends on the very premises that make the argument circular, and if the argument is admittedly vicious precisely to the extent to which it is circular, then that viciousness is no small matter.

There is no good reason, as far as we have seen, to think that the size of the circle in any way mitigates the viciousness of epistemic circularity. It might be a popular notion to think so, but it is apparently groundless.

In this chapter, the thrust of my argument has been that meta-coherentism posits the existence of epistemically circular arguments. But such circularity is vicious, as was argued at length in Chapter 2. As far as we have been able to ascertain, the meta-coherentist lacks any grounds on which to claim that the viciousness of that epistemic circularity is, somehow, lessened by adopting his position. Neither the Method of Reflective Equilibrium (and Rescher's "instrumental justification") nor the claim that the size of the circle matters, have been shown to render the arguments from Chapter 2 any less forceful. Indeed, after close examination of the meta-coherentist's defense of the claim that epistemic circularity can be virtuous, we have found that this defense rests on either bald assertions or else confusions.

V. Meta-Regressism.

The Meta-Regress Argument, like the ordinary regress argument, permits a position according to which a belief (in a j-standard, in this case) is justified by an infinite chain of other beliefs. In Chapter 1 this position was introduced and called meta-regressism . In particular, the view is that, for some j-standard, so-and-so accepts (dispositionally) an infinite series of standards, or that "there is" an infinite series of them, supporting but not containing the standard. (`To support', you will recall, in this context means `to license belief in a premises or premises of an argument that justifies belief in a conclusion'.)

The view that infinite chains of dispositional beliefs can justify anything seems implausible on its face and has gained only some (only a very few, in print) adherents. So there is, I think, no need for us here to rehearse the half-dozen or more arguments on either side of the issue. The following more modest discussion ought to suffice for the bare purpose of putting meta-regressism to rest.

We are considering meta-regressism as a possible solution to the PMJ, which, as we concluded in Chapter 1, is essentially the problem of offering a satisfactory argument for a j-standard. On first glance, to countenance meta-regressism as a strategy for solving this problem seems tantamount to claiming that one could, somehow, produce an argument with an infinite number of steps.

But, of course, such a thing can be described logically. One could very easily describe how the soundness proof of a deductive system could be iterated ad infinitum. The description might be reworked and reinterpreted as an argument for the reliability of deduction as a belief-forming procedure (call this `DDP' for `deductive doxastic practice'). Then one claims, of the proof thus indicated, that it is what provides adequate justification of DDP.

This raises an important question, however. Is it the proof itself (something abstract and merely indicated, never described in all its detail) or else the description of the proof (something spoken or written down, or perhaps just entertained mentally) that is claimed to justify DDP? What we are presented with is not the proof itself, because that's infinite, and we are finite creatures. What we get is a description of the proof – which, perhaps, we might also call "the proof," but speaking more loosely.

As we epistemologists are examining the description of the infinitely long proof, hoping for a justification of DDP, it is evidently not this indicated proof that directly justifies our belief. If anything, as far as we finite humans are concerned, it is the description of the proof, or rather, some belief about a described proof, that does any direct justifying. If we are persuaded of anything, it will be that (1) this infinitely long proof exists, and that (2) it, an abstract object, is adequate to support DDP. Fortunately, neither (1) nor (2) requires that we have an infinite number of beliefs, dispositionally or otherwise. But, unfortunately for hopes we might have had for meta-regressism, we are now proposing a different solution to the PMJ; our new proposal could be any number of other solutions, e.g., some variety of meta-foundationalism. The fact that some premise in a solution to the PMJ makes reference to an abstract proof with an infinite number of steps hardly commits one to meta-regressism.

My claim is that the following is always going to be the case with an alleged meta-regressism: the infinite regress (proof) is described, but then evidently it is not any infinite series of beliefs that is doing the justification, but rather the less ambitious project of believing that this proof exists and is adequate to its purpose. This claim of mine is an empirical observation, open to falsification, but, I think, clearly true. Anyone we see busily engaged in rehearsing as many steps of an infinite proof as he could before he died would perhaps be said to have some extra evidence that the proof, an abstract object, exists – but even that's highly doubtful, of course. The proof rehearser would not, of course, be said to believe the conclusion of the proof just on the basis of having believed many of its premises. This is not how such proofs are used by human beings, and it could not be otherwise with us.

If one objects that we could have an infinite number of dispositional beliefs – a question I won't comment on and thus that it is still quite possible that beliefs in all of the premises of an infinite proof are what justify belief in the conclusion, I offer the following reply. The relevant question appears to be an empirical one: if an infinitely long proof is in fact used to justify my belief in its conclusion, what best accounts for that doxastic justification? Is it my single, occurrent belief that the proof exists, or is it instead an infinite number of dispositional beliefs, most of which I am unaware of having? Economy alone dictates that the former answer is correct.

Considerations of economy generally dictate that meta-regressism is a theoretical option open only to God, who is up to the job of believing an infinite number of things. In accounting for human justification, if a believer makes reference to infinitely long proofs, we do not, on account of that, have to say that the believer is committed to meta-regressism. Hence, alleged meta-regressists can be neatly dealt with under the other options listed in the Meta-Regress Argument. Most of them, I speculate, would be properly classified as self-supportists, and in that case the arguments in Chapter 2 would be brought to bear.

Now we abruptly switch gears.

VI. The Problem of the Criterion.

The ancient Problem of the Criterion (POC ), introduced by Sextus Empiricus,[133] and the modern variety due to Roderick Chisholm,[134] have a great deal in common with the Problem of Meta-Justification. Indeed, the three positions that Chisholm outlined on the problem, viz., methodism, particularism, and skepticism, are each positions one might take on the PMJ. It will be instructive, before we evaluate methodism and particularism, to introduce the POC in some more detail. We will focus on Chisholm's formulation of the problem.[135]

It seems to me I know quite a lot about my immediate environment; for example, that the light is off in this room, that music is playing, and that there are birds chirping away outside my window. But, Chisholm says, such knowledge-claims about commonplace facts open one up to skeptical inquiry: how do I know that birds are chirping away outside my window? I might supply a general reason, or criterion, according to which I know: if it seems to me that there are sounds, etc., in ordinary conditions, etc., then I know that there are birds, etc. But again the skeptical inquiry can be pressed: how do I know that this criterion is true? Here, Chisholm says, it is natural to support the criterion by referring to many individual instances of knowledge. Indeed, it is obvious that I do know about the birds, whether the light is on or off, and many other facts about my immediate environment. We may construct (and justify) a criterion of knowledge based on a lot of individual cases.

But here the skeptic appears to triumph. He says, consider these two questions you are attempting to answer:

(1) What do I know? What is the extent of my knowledge?

(2) How do I know it? According to what criteria can I be properly said to have knowledge?

"Observe," the skeptic says, "that you justified your answer to (1) with your answer to (2); and then you justified your answer to (2) by referring back to your answer (1). You are arguing in a circle. And indeed it couldn't be any other way. You cannot answer (1) without first having answered (2), and you cannot answer (2) without first having answered (1). Since you cannot answer either without begging the question, you lack both knowledge and any reliable criteria according to which you might show that you have it."

There are, according to Chisholm, two possible replies to the skeptic. One might begin with an answer to (1), specifying several instances of knowledge; the procedure then is to formulate a criterion in accordance with those instances. This approach is called particularism .[136] On the other hand, one might begin with an answer to (2), specifying some criterion of knowledge that one antecedently regards as certain, or at least sufficiently plausible; then one decides on the epistemic status of particular beliefs by reference to that criterion. This is the approach of methodism .[137]

Chisholm is a particularist, but he nonetheless holds that any particular position on the POC "begs the question" against the other positions.[138] The particularist simply takes his knowledge-instances for granted, "begging the question" against the methodist who believes a criterion is needed to identify the instances; the methodist takes his criterion for granted, "begging the question" against the particularist who insists that some support for the criterion is needed (inductive support by instances, in particular). Of course, none of this implies that those espousing either one of the positions must beg the question in the precise sense of having no arguments for the position itself (e.g., particularism ), or against the contrary position (e.g., methodism ). That is, what are taken for granted are not particularism and methodism themselves, but instead the criterion and knowledge-instances that the doctrines are about.

Recall what the meta-justification problem asks: for any given standard of justification, what is its justification? Of course, the word `criterion' as used in the Problem of the Criterion is another one of many names for a broad category of claims I refer to as "justification standards" (or even more broadly, "epistemic standards"). So the particularist, at any rate, is concerned to solve the PMJ.

A particularist's argument for his criterion is inductive. The procedure is to gather a lot of clear instances of knowledge, and then devise a criterion that best explains why they are all instances of knowledge.[139] When the criterion is settled upon and the argument is written down, it will closely resemble the sort of track record argument described in Chapter 1 – something like this:

(1a) S's belief that p1 meets conditions c at time t1.

(1b) S knows that p1 at t1.

(2a) S's belief that p2 meets conditions c at time t2.

(2b) S knows that p2 at t2.

(na) S's belief that pn meets conditions c at time tn.

(nb) S knows that pn at tn.

(C) Generally, if S's belief that p meets conditions c at t, then S knows that p at t.

The hard abductive work goes into formulating "conditions c" in a precise enough way that for all of the (nb)'s (which are taken for granted), their corresponding (na)'s are true. The inductive formulation of the argument above just tabulates the work.

Consequently, at first glance, the particularist looks like a self-supportist – recall that self-supportists take track record arguments to be adequate to support j-standards. Belief in each (nb) is said to be justified in accordance with (something close to) the conclusion. But let us have a closer look at the situation.

Essentially, the particularist claims that he can fix on particular instances of knowledge without, antecedently, having a criterion to decide that they are instances of knowledge. In the track record argument above, the particularist would say that each (nb) is true, and that he needs no criterion – or other epistemic standard – to decide that it is true. He takes them all for granted, and is happy to do so. So now the particularist looks less like a self-supportist; he appears to want to claim that it is neither the conclusion (the criterion he arrives at) nor any other epistemic standard that justifies his knowledge-claims in his argument's premises.

This discussion makes a certain confusion plain. The particularist might not have arrived at his instances of knowledge by the use of a criterion; but does he claim that those instances are, nonetheless, instances of belief that are justified in accordance with a standard? Or are we to take his disavowals of the use of a criterion as an indication that he also denies that the beliefs he claims to be knowledge are justified in accordance with any standard?

The particularist might, when confronted with this confusion, wish to clarify his position by saying that his knowledge-claims are indeed justified in accordance with some standard or other. Perhaps he will explain that particularism is a philosophical method that does not commit him to a meta-epistemological position, and that his meta-epistemological position is, as it turns out, (e.g.) self-supportism. If that's the case, we may confront the particularist with the arguments from Chapter 2, and the issue is finished.

But the particularist might, on the other hand, claim that his position is quite distinct from self-supportism. On this variation, his position is indeed that there are some knowledge-claims that not only may he assert without support, but that are justified without there being any standard according to which they are justified. It is possible that certain beliefs are justified, but not justified in accordance with any standard of justification; and the knowledge-claims he takes for granted, in arguing for a criterion, are indeed such justified beliefs. In that case, he claims that his argument for the criterion lacks any licensing standard, to use the terminology introduced in Chapter 1, because the premises lack a licensing standard.

The latter variety of particularism is what we will examine in this chapter; let us call it strict particularism.

It is not hard to understand that a similar confusion can arise for the methodist. That is, when the methodist claims that he knows that a certain criterion is correct, are we to say that there is some further criterion in accordance with which his belief in the criterion is constituted as knowledge? If so, the methodist might well espouse meta-coherentism or some other meta-epistemological position, while maintaining that methodism is just a philosophical method. But if not – if the methodist says there is no further criterion – then he shares with the strict particularist the view that certain beliefs are correctly regarded as justified, but not justified in accordance with any standard of justification. And then we will call his view strict methodism.

The strict varieties of particularism and methodism each combine similar meta-epistemological views with different methodologies. What we will examine next is the meta-epistemological view they have in common.

VII. Unlicensed Meta-Epistemology.

Strict particularism and strict methodism make a similar proposal. The proposal is that there are some epistemic claims – claims that centrally involve epistemic terms such as `knowledge', `justification', and `evidence' – that are true, but for which there is no licensing standard. And a licensing standard of an epistemic claim is, as discussed in Chapter 1, an epistemic standard that best explains the fact that the epistemic term correctly describes what the claim alleges it to describe.

As an example, let's take the classic claim of a classic particularist. "Here is one hand," said G. E. Moore, holding up a hand.[140] We may attribute the following epistemic claim to Moore as well: "I know that this is a human hand." An epistemic standard would be a licensing standard for this claim, if that standard best explains why Moore was correct to say "I know this is a human hand." We need not attempt to formulate the standard precisely, but a rough example would be something like this: "If it seems to someone that he is clearly seeing something just a foot or two away from him in good light, and it appears to him that he is fully awake, sober, etc., and he has no reason to think his senses are deceiving him, then he knows that he is clearly seeing that thing."[141]

Our claim, then, would be that this standard best explains the fact that Moore knows that this is a human hand. The standard provides a sufficient condition for knowledge that applies to many different cases, including the present case; and not only does the obtaining of the condition imply that Moore knows, it explains that he knows. In the example, we might say, it explains this fact at least as well as any other standard can. The Holy Grail of recent epistemology has been to formulate epistemic standards that succeed in this way – that explain epistemic facts (if you will) better than, or at least as well as, any other contenders. The standards that succeed in this are the standards, as I say, that "license" a belief's claim to justification (thus the usefulness of the word `license').

Consider what the strict particularist might say of this case: there is no such explanation. It is certainly a fact that Moore knows that this is a human hand – that's granted – but there is no standard to appeal to, that explains why Moore knows in this instance.

But a more clever particularist might present his position as follows: "There is no criterion that does the explanation – that's my essential position, after all. But that hardly means there's no explanation at all for this instance of Moore's knowledge. Obviously, he knows it is a hand because he sees it in good light, etc. The point is that there is no generalization – `covering law,' or criterion, or j-standard, or whatever you'd like to call it – that contributes to the explanation."

These two contrasting defenses naturally lead one to the question whether strict particularism is indeed committed to the position that there is no explanation whatsoever for fundamental knowledge-claims; the defense just presented answers the question in the negative. Thus far, admittedly, the strict particularist has been described only as lacking a criterion that accounts for particular knowledge-instances; admittedly, it is a logical possibility that there could be such an explanation without a covering criterion. After all, there have been views of causal explanation, e.g., Wesley Salmon's, that reject the "covering law" model.

But the sort of explanation needed here is semantic explanation: we are looking for an account, even a very partial account, of the meaning of `know' that explains why this application of the word is correct. However matters of causal explanation might be, semantic explanation does clearly require the application of a criterion, or generalization of some sort – a universally quantified conditional claim.

If that's correct, then the strict particularist is committed to the claim that there's no explanation for the fact that `know' applies in Moore's case. Or, at least, there is no adequate explanation, because an adequate explanation would require the application of something like a criterion.[142] He just does know it.

Moreover – to return to the main thread of the discussion – the strict particularist collects together a number of such (unexplained) epistemic claims, and uses them to arrive at a criterion (an epistemic standard). That's his basic procedure.

Similarly, the strict methodist begins with a standard that he claims to know (or to be justified in believing), and about which he says, essentially, there is no explanation of the fact that he knows (is justified in believing) it. Moreover, he uses this standard to arrive at a number of epistemic claims.

Plainly, then, one way to evaluate the merits of strict particularism and of strict methodism would be to decide whether any epistemic claim, whether about epistemic standards or about other, more particular beliefs, could possibly lack an explanation for the fact that the epistemic terms it contains are correctly applied.

There are some terms that lack such explanations, of course – terms that are semantically primitive , we might say. For example, an ontologist might suggest that `object' is a term such that there is no (informative) way to account for the fact that it does correctly apply, when it does; others might suggest, instead, that `property' and `relation' are such terms. In systems of logic, two connectives are often presented without explanation; one might well say, for example, that the occurrence of `and' in a sentence lacks any explanation, which hardly keeps sentences that employ it from being true.

Presently at issue is whether there are some epistemic terms, such as `know' and `justified', that are semantically primitive in this way. Or rather, the issue is whether they are primitive in some uses, but not in others; after all, both the methodist and the particularist are, of course, busily formulating adequate explanations for the application of epistemic terms. The particularist, for example, takes a lot of epistemic claims for granted and uses them to fix on a criterion, i.e., a statement that explains when an epistemic term such as `know' is correctly applied.

Why think, then, that some uses of epistemic terms are semantically primitive? More precisely, why think that the uses of `know' in the (strict ) methodist's claim to know that his criterion is correct, or in the (strict ) particularist's allegedly clear cases of knowledge, are semantically primitive? The formulation of the POC makes it clear enough: only if those uses of epistemic terms are semantically primitive can we escape "the wheel" of epistemic circularity (and hence, allegedly, skepticism ). This accounts for why methodists and particularists do in fact regard some uses of epistemic terms as semantically primitive.

But the challenge remains: can they give some other, non-question-begging reason for thinking that those uses of epistemic terms are semantically primitive? Very likely the answer is in the negative, because on the face of it, to suppose just those uses of epistemic terms to be semantically primitive appears ad hoc. In addition, consider the following two arguments; the essential point in both is that strict methodists and particularists, to avoid epistemic circularity, must regard as semantically primitive more than they were prepared to admit.

First, on particularism. As we have seen, the strict particularist constructs an argument that greatly resembles the track record arguments presented in Chapter 1. And, as we saw in Chapter 1, there are a number of conditions on using an argument to get a justified belief in an argument's conclusion: not only must one believe the premises justifiedly, one must, in addition, believe justifiedly that the conclusion follows from the premises. The strict particularist insists that he knows the premises without justifiers, or anything else that would explain or account for how he knows them. But we may ask him: how do you know that the conclusion, the criterion you generate from instances of knowledge, follows from the premises? By some sort of induction, evidently. But what licenses you in making the inference?

Here the particularist faces a nasty dilemma. On the one hand, he might reply that there is some epistemic rule regarding which inductions are justified, and that this particular induction follows that rule. But that puts him back "on the wheel," of course: even if he claims that he need not know what the rule is, the arguments surrounding that defense from Chapter 2 are easily brought to bear. On the other hand, he might say that this particular induction is justified, but there is no licensing rule, nothing that accounts for its justification. And here again we may say that his solution appears ad hoc. If he is permitted to say that this inference is justified without some covering explanation, then why is he not permitted to assume that virtually any inference is similarly justified?

Second, on methodism. The argument here is much the same. The methodist claims to take a certain criterion for granted. But in addition, like the particularist, he must take for granted whatever inference rule he uses (in applying the criterion). Since he is already assuming one "criterion," what's another to him?

But worse than that – supposing that his criterion is a conditional claim of the form, "If S satisfies conditions c, then S knows that p" – he must take for granted all of the instances of the antecedents of his criterion. For example, a methodist Cartesian might assert: "If I clearly and distinctly perceive that p, then I know that p." For every p that the Cartesian wants to say he knows, he must take for granted that he is justified in believing – without any explanation – that he clearly and distinctly perceives that p.[143]

And now the methodist faces the same dilemma that the particularist faces. Either he relents and says that there is some way to account for the justification of his inferences and for the antecedent-instances of his criterion, in which case he is subject to charges of epistemic circularity; or else he insists that these particular uses of the epistemic terms are, again, primitive, in which case his position seems increasingly ad hoc and thus untenable.

The strict particularist and methodist might object that everyone must take some terms as primitive, after all. So why should it be a problem that they take certain epistemic terms as primitive? Why should there be any requirement to explain this at all – so that charges of ad hoc positing of primitive terms have any negative force at all?

To be clear, the problem is not merely that the particularist and methodist must take certain uses of epistemic terms to be primitive. The problem, properly understood, lies in the fact that the particular use of the terms that are taken to be primitive are so taken precisely in order to solve a philosophical problem, the POC. But the fact that taking just these uses of these terms does allow one to solve the POC does not appear to be anything like a good reason to regard just them as primitive.

While, granted, everyone must take some terms as primitive, it does not follow from that that the choice of terms is arbitrary, or that terms may be chosen simply in order that problems can be solved handily. Presumably, there are reasons why primitive terms ought, in philosophy, to be taken as primitive,[144] just as there are reasons why certain propositions may be taken as philosophical "starting points" while others may not.

If the explanation of why a philosopher takes some, but not all, uses of some epistemic terms to be primitive is that doing so permits one to solve the POC, one can very justly accuse him of doing something ad hoc. The fact that taking these terms to be primitive allows one to solve the POC is not a reason to think that the terms ought to be taken as primitive. This is a philosophical error on the order of taking as an axiom a proposition that nearly everyone believes to require and to be amenable to argument. The fact that doing so might handily avoid some philosophical problem is not a reason to think the proposition ought to be taken as an axiom.

VIII. Strangely Justified Beliefs.

The reader might think we have been a bit too quick in dismissing particularism and methodism. Perhaps there is something to what they, in their strict varieties, say. One way to regard their shared assumption is that there are certain claims that are justified without justifiers, i.e., without anything doing the justification. Surely there is some (perhaps minimal prima facie) plausibility to the odd notion that a belief might be justified – that one ought to believe it, that it has some claim to being true – in spite of the fact there is nothing "doing the justification." And if such an animal might exist, then surely the strict particularist or strict methodist might have correctly identified these strangely justified beliefs. (For convenience, let's refer to these beliefs as indeed strangely justified.)

Foundationalists might be initially attracted to the view that strangely justified beliefs exist, since they are committed to the existence of properly basic beliefs – i.e., those that are justified, but not justified by other beliefs. But notice now that properly basic beliefs are not strangely justified. The typical foundationalist claims that there are indeed justifiers for basic beliefs. Typically, the justifiers are mental states such as perceptual appearances and internal sensations; accordingly, there are conditions for the application of `justified' as applied to basic beliefs. A strangely justified belief, by contrast, is not justified by anything at all; nothing like mental states account for its justification, because nothing at all accounts for its justification.

I suspect that foundationalists, when properly inducted into these mysteries, will want to have nothing to do with strangely justified beliefs.

Notice too that strangely justified beliefs are not what have been called "blind posits," i.e., beliefs that are simply taken for granted, or for which no claim about their justification is made. Our strict methodist does wish to claim that his ultimate criterion is justified. If he did not make that claim, methodism could scarcely be called a solution to the POC. It would be better regarded as a variety of skepticism, since it is the skeptic, after all, who essentially makes the claim that neither knowledge claims nor criteria for knowledge can be justified.

Strangely justified beliefs thus compose an epistemic category that is, today, not well known or commonly considered. One might be apt to add that strangely justified beliefs have had little currency in the history of philosophy; but this is, on reflection, not so clear. It was not until the twentieth century that the terminology of "justification," "warrant," and the like became widespread. Earlier discussions of foundations tended to be couched in language that was not so closely tied to notions of justification, e.g., `first principles', `axioms', and `maxim'. It would be a matter of some historical debate whether, just to take one example, Locke's "maxims" (or, what might be the same thing, beliefs in maxims) are best regarded as properly basic beliefs, strangely justified beliefs, or blind posits.[145] (Possibly the historic question would be best left unresolved.)

However that is, I suppose the main reason that, at present, strangely justified beliefs are not widely discussed by epistemologists is expressed in the following line of reasoning. Strange justification is justification without justifiers. If I say that "This is a hand" is justified strangely, my claim is that "This is a hand" is justified, but that nothing at all does the justifying. There is nothing that stands in some relation to the belief, in virtue of which the belief is justified. But, I suppose, it is so common to think that justification does absolutely require justifiers, that this requirement is hardly ever given expression. It seems essential to the very notion of justification that there be justifiers. Indeed, one ordinary (mainly non-philosophical) use of the word `justification' is practically synonymous with `reasons' or `evidence' or `argument' – i.e., with the justifiers themselves, and not with any quality that a belief might have thanks to support by the justifiers. So we can take it that there simply is not any legitimate sense of `justification' according to which the use of `strangely justified' can be countenanced at all.

This line of reasoning is admittedly tenuous. But, given the very slight plausibility of the existence of strange justification to begin with, a stronger argument does not seem necessary.

Insofar as strict methodism and particularism are committed to the existence of strangely justified beliefs, then, that is another objection to those theories. Considering, as well, the objection that the theories require the ad hoc positing of certain uses of epistemic terms as primitive, we may put strict methodism and particularism to rest. Moreover, the looser versions of methodism and particularism are, as we said, in principle indistinguishable from other proposed solutions to the PMJ, which have been refuted in Chapter 2 and earlier in this chapter. So we must look elsewhere for a solution to the POC and the PMJ – or else accept a skeptical result.

IX. Conclusion.

It is time to put this chapter's work in context. We have, essentially, eliminated a series of proposed solutions to the Problem of Meta-Justification, thus expanding the Meta-Regress Argument presented in Chapter 1.

In Chapter 2, we found that self-supportism must be rejected: j-standards cannot support themselves without running afoul of a vicious epistemic circularity. And, as we found early in the present chapter, meta-coherentism doesn't make the situation any better by positing a circle or loop of standards; that proposal too runs afoul of epistemic circularity. It does not help to suggest that the process of developing a coherent system does not intrinsically involve arguing in a circle (which is just irrelevant to the question of epistemic justification). Moreover, mentioning the size of system in which the circle is embedded does not help to mitigate this circularity in any discernible way.

Another logical possibility is meta-regressism – that there is an infinite series of j-standards, with each j-standard the conclusion of an argument that is licensed by another j-standard that is itself the conclusion of another argument, and so on ad infinitum. But, however things might be for God, the most parsimonious account of how human beings might use such an infinite series of arguments to arrive at a justified belief is not to suppose that we have an infinite number of dispositional beliefs, but rather that we have a belief in the existence of an infinitely long argument, and this latter belief is what does any justification. Any proposal resembling bona fide meta-regressism will, thus, reduce to some other solution.

Next we considered strict particularism, the view that j-standards are justified by premises about particular instances of knowledge or justified belief – but there are no criteria by which the application of `know' or `justified' to those premises can be explained. But that `know' or `justified' do apply here without explanation (i.e., that particular instances of the use of these terms are semantically primitive ) to just these premises appears to be simply ad hoc. Worse, at least on some versions of strict particularism, there are no justifiers for the premises. But that there could be justified beliefs without justifiers is extremely implausible.

So we might fall back to strict methodism, the view that j-standards are justified but that there are no instances of justified belief and no other standards that justify them. This view faces the same problems that strict particularism faces: it involves ad hoc reasoning and appears to be committed to the extremely implausible view that "strangely justified beliefs" exist.

Are we left, then, with the view that j-standards are not, ultimately, justified at all – that the best we do is to make some "blind posits"? Isn't that tantamount simply to skepticism? These questions will be explored in the next chapter.