Symposium Introduction

Coherence and Justification: Ted Poston’s Reason and Explanation

1. Reasons and Coherentism

The coherence view of justification is, in an important way, a dissolution of an old philosophical puzzle, the regress problem. Every inquisitive child seems to know how to get the regress of reasons going—you just ask why, and a reason is given. But reasons can be challenged, and if they are to play their role as reason for what they are given, they must survive challenges. And so more reasons are given, and more why questions may be asked yet of them. This pattern delights not only children, but skeptics, too. This is because it seems that the three options for adequate answers to why questions to proceed all seem to be empty of what complete reasons need. Clearly, reasons cannot go on to infinity. After all, as Aristotle observed, nature flees the infinite. Another option is the circle, but this instantiates the problem of begging the question. And so our model for reason-giving cannot also share the same form as a common fallacy. Finally, there is the thought that there are some self-intimating or self-evident propositions, ones from which the rest of our reasons derive their support. A foundation, as it were, a bedrock of justification. Alas, this project, too, has the problem of identifying why these, instead of those, foundations obtain, so the project of providing reasons seems far from complete, and there is the lingering puzzle of how we have could access to such important truths.

From this skeptical mire, the coherentist program arises. The problem, the coherentist points out, is that all the models for reasons that run the regress have been in error—primarily, the model of serial justification. For sure we speak our reasons one at a time, but that is only because we can say only one sentence at a time. That looks like the reasons come in serial, or linear, fashion. But this is only an illusion—they are spoken one at a time, but they function as groups, as systematic wholes. It is because we know many things that our reasons hang together. Only when we understand a wide range of things can we reason reliably in the relevant domains. And so, the regress problem is dissolved by coherentist holism—there is no chain of reasons to complete or loop around on itself or extend to the horizon. Rather, there are the beliefs and experiences we have, and they come as broad stories and packages for the context of every why question. These frameworks of beliefs and experiences in the background make the tidy linear reasons we happen to give relevant. For every domain of Why? questions, there is a framework of experiences and assumptions we must bring to bear on the question; thereby, things fit together, make sense, and amount to a whole we can understand. That’s the coherence theory.

2. Poston’s Conservative Explanationist Coherentism

Ted Poston’s new account of coherentism proceeds along two substantive lines. This is because coherentists regularly labor to answer two challenges. The first challenge is to articulate how coherence with a group of background beliefs can confer justification. It seems that it may give an exculpation for (or for others a prediction of) a subject’s commitment, that it is the kind of thing someone with those beliefs would reasonably take true under these conditions, but it doesn’t seem on its face like a justification. This is because the background beliefs themselves should be justified, otherwise the question is how they, without justification, can justify. Coherence looks at best a way justification can be transmitted or enhanced, but it does not seem to be a source or ground for justification.

The second issue is connected to the first, which is what exactly coherence is. It must be a stronger relation than mere consistency (where it is possible for all the propositions to be true), but it cannot be so strong as deductive entailment. And once we are clear as to what it is, the first question returns—how is this relation more than merely a means of transmitting already existing justification?

Poston’s program of answering these challenges is that of defending conservatism and explanationism as components of his coherentism. Poston defends a form of epistemic conservatism wherein, for propositions we survey for justification, “belief . . . confers some positive merit on the proposition” (2014: 18). So it is in believing the coherent set of propositions that a subject has the initial reason to believe any member of the coherent set. As Poston notes, “Unless the data used for coherence reasoning has some initial probability, mere coherence cannot raise the probability of the belief” (2015: 43). Coherentism works only if conservatism works. Conservatism is, at least on first impression, an implausible theory, as it seems that one could say that one’s reason for believing, say, that the cat is on the mat is that one, in fact, believes that the cat is on the mat. But Poston makes the case that the implausibility here is another illusion of reason-giving, as opposed to the state of being justified. One’s conservative reasons being unassertable (perhaps because they come as large packages or because they are hinge propositions in one’s background beliefs), is not yet a case that they aren’t reasons.

The second constructive edge of Poston’s coherentist program is his explanationist model for epistemic justification. Poston’s core claims are that all justified ampliative inference is fundamentally explanatory. Our inferential practices are not straight-rule procedures, but rather cases of plausibility-assessments of options. In order to weigh the value of the competing options, we must appeal to their explanatory virtues in light of many other things we know. Consequently, Poston holds, “the property of being a reason is a property a single proposition has only in relation to a background body of beliefs” (2015: 57).

Poston is, on analogy with Williamson-style Knowledge-First epistemologists, an Explanation-First epistemologist. Knowledge-First epistemologists hold that knowledge is a kind of cognitive simple—one does not define knowledge by breaking it down into constituent parts. Poston is an Explanation-First epistemologist; for him, explanation is a kind of cognitive simple. First, the explanation-relation is primitive in that it lacks a metaphysical analysis. Second, the cognitive state picked out by understanding “because” statements is itself also a cognitive simple. So Poston is a kind of two-ply Explanation-Firster—explanation is both metaphysically and cognitively simple. He observes that “explanation is a basic human activity” (2014: 76), one that we master prior to many of the other notions we currently try to analyze it into. Given that Poston holds that the core of epistemic justification is rooted in the comparison of explanations, there must be more features of explanation that are gradable and comparable. These are conservatism, simplicity, and explanatory power. And so, the less an explanation damages our background beliefs, the more wide-ranging the explanations it provides, and the more parsimonious and elegant it is compared to its competitors, the better the explanation. Given these tools, Poston states his Explanationist Theory of Justification (Ex-J):

S has justification for believing p if p is a member of a sufficiently virtuous explanatory system E, and e is more virtuous than any (p-relevant) competing system, E’ (2014: 90).

Poston makes the case that Ex-J is a mentalist and evidentialist view, as justification for subjects depends on what evidence they have, and what evidence they have depends on what mental states and conditions they are in. Changes in what evidence a subject has are always changes in what mental states the subject is in.

The upshot is that Poston has proposed a framework view of reasons that is a non-skeptical holist internalism. It is a form of internalism, in that Poston requires that all the contents that comprise the justification are ones internal to the subject being evaluated; it is holist, in that justification is conferred on systems; and it is non-skeptical, in that it is designed to maintain many of our ordinary commitments about the world and our sensory access to it, the progress of science, and the truths of logic as justified. Poston’s book is an exercise in coherentist epistemic optimism.

3. The Responses

Coherentism, given its dialectical position with regard to a perennial problem in philosophy, has a number of articulations posed and challenges antiposed. This is also the case with Poston’s new model—as the view is reformulated and updated, so may the challenges be.

Coherentism has regularly had the problem that alternate systems dog it from the earliest of its articulations; the commitment of having only relations between mental states being the prime feature of justification yields the spectre of those well-formed systems spinning off with nary a connection to the truths they purport to be about. Thomas Dabay poses a version of the challenge and indexes it to the sources of initial justification for Poston’s coherentism, the hinge propositions that are conservatively justified. Dabay’s challenge is a trilemma for Poston. In the face of a competing minimally virtuous system, Poston must either: (a) reassert some part of the preferred system, and thereby embrace a form of dogmatism; (b) hold that there must be some justification in addition to the conservative that favors one over the other, and thereby evacuate the conservatism; or (c) accept that one cannot hold one is superior to the other, and thereby accept skepticism.

Ian Schnee, also, challenges Poston’s conservatism. Schnee provides four lines of criticism. First, Schnee takes it that Poston’s restriction of justification for mere beliefs only to empty symmetrical evidential conditions severely reduces the number of propositions that can do the work for the conservative-coherentist program. Second, Schnee worries that Poston’s restriction seems ad hoc in a way that undercuts the motivations for the program. Poston claims that the program is evidentialist in spirit and aspiration (particularly in the case against foundationalism), but his conservatism is in conflict with evidentialism. Third, Schnee challenges Poston’s defense against the objection that the conservative view would have to accept claims of justification of the sort: “I believe that p, and my justification for it is that I believe it.” Poston’s defense is that what’s wrong about the statement is that it isn’t pragmatically appropriate, because it fails relevance. Schnee argues that this strategy is devastating, since the justification that survives, assuming that it fails relevance, is itself not connected to truth. Fourth, and finally, Schnee takes issue with Poston’s claim that his conservatism meshes with the first-person perspective on justification. Schnee charges Poston with having confused the issue of taking a subject’s perspective into consideration with taking the subject’s views about the perspective as relevant to justification.

Georgi Gardiner argues that Poston’s epistemic conservatism is in tension with his coherentism. Gardiner first argues that the view would have the implausible outcome that conservatism would let a mere belief be justified, but once, say, meager evidence comes in, the belief would then be no longer justified. Gardiner’s second argument is that Poston’s speech-act theory solution to the problem of holding one’s mere belief as justification for the belief doesn’t work for cases where the claim is merely thought, instead of asserted. Third, and finally, Gardiner challenges the idea that evidence can be both symmetrical and empty. As a consequence, Poston’s target class for conservative belief is not possible, and so of no use for a coherentist theory of justification.

Andrew Cling continues the challenges to Poston’s conservatism. Cling’s challenge is that if one is both an evidentialist in holding that justification requires that one must have evidence available and an epistemic conservative in holding that one’s beliefs come with epistemic credibility in virtue of their being believed, then one is caught in an epistemic impasse. The impasse shows that Poston, in the end, cannot claim his view is robustly anti-skeptical. Cling shows how Poston’s conservatism arises from his blocking the regress problem for justification in that Poston must deny that all justified beliefs require reasons. So Poston must deny that any epistemic evaluative difference between two subjects’ beliefs is grounded in the evidence they have. But Poston claims that his conservatism implies that, in some cases, the difference is just in the believing. This, Cling argues, makes Poston’s view not so different from a mitigated skepticism, as it seems that Poston takes it that one can be justified in believing a proposition even if, from one’s own perspective, it is not clear what makes it epistemically rational to do so.



Conservatism Alternate Systems and a Trilemma for Poston

1. Introduction

In this response to Ted Poston’s Reason and Explanation, I critique his conservatist thesis that a person’s believing some propositional content “confers some positive epistemic merit on the content of belief.” (19) To this end, I begin by outlining the relevant aspects of Poston’s conservatism in §2 before posing a version of the alternate systems objection to Poston in §3. Finally, in §4 I use this objection to motivate a trilemma for Poston, the horns of which lead to dogmatism, self-defeat, and skepticism.

2. Poston on Basic Reasons and Empty Symmetrical Evidence

Before I begin my critique of Poston’s conservatism, I must first commend him on his framework view of reasons. Chapter 3 of Reason and Explanation, where Poston develops this view, is a masterful defense of coherentism against a perennial worry, and anyone interested in coherentism, foundationalism, or related issues would be well served to give this chapter the close reading it deserves. Poston formulates this worry in terms of what he calls the Basic Reasons Dilemma: any epistemological view is committed either to the existence of basic reasons—i.e., non-inferentially justified reasons that support other propositions—or to the non-existence of basic reasons. If a view is committed to the existence of basic reasons, then it falls prey to Sellars’s Myth of the Given and other anti-foundational arguments. But if a view is committed to the non-existence of basic reasons, then the view is viciously circular. This is because the purpose of basic reasons is to end explanatory regresses, and without them the only way to end such regresses is through circular reasoning. Either horn of the dilemma leads to an objection, and the worry is that this is the end of the story (46–49).

Poston argues that he can escape this dilemma, given how he develops his conservatism in chapter 2. First, a conservatively justified reason is justified non-inferentially, so it can serve the regress-ending role that is associated with basic beliefs. And second, such a conservatively justified reason cannot be used to support other propositions, so it is not itself a basic belief. To my mind, Poston’s ability to hold together these two seemingly contrary points is his greatest achievement in Reason and Explanation, and I direct curious readers to §3.3 of Poston’s book to see how he pulls it off (56–61).

However, the ultimate success of Poston’s argument at §3.3 is contingent upon the success of his earlier argument for conservatism from chapter 2, and it is here where I see room for critical pushback.

Conservatism is the claim “that belief confers some positive epistemic merit on the content of belief.” (19) Poston rightly points out that this is a controversial claim, and sets out to limit its application to only those cases where it is plausible. Towards this end, we must distinguish two types of case from each other. First, there are cases where the evidence for a believer’s belief in p is asymmetrical in the sense that the believer’s evidence for p is not equal to her evidence for not-p. Conservatism in asymmetrical cases is implausible: if the believer has more evidence for p, then it is that evidence and not her believing p that justifies her belief in p; and if the believer has more evidence for not-p, then her belief in p is not epistemically meritorious as conservatism would imply, but simply irrational. For this reason, Poston limits his discussion of conservatism to cases of the second, symmetrical type where the believer has equal evidence for p and not-p.

However, even in symmetrical cases conservatism seems questionable. When a believer has equal evidence for and against p, the proper response would seem to be for her to suspend belief. To do as conservatism implies and maintain her belief in p on the force of her original belief in p seems dogmatic and intellectually irresponsible. Poston responds to this concern by distinguishing between two subtypes of symmetrical case—one where the symmetry is the result of there being different pieces of evidence that counterbalance one another, and the other where the symmetry is the result of there being no evidence one way or another. Poston says little about the former cases, leaving the reader to conclude that suspension is the proper response. However, he argues that in the latter cases—cases of empty symmetrical evidence—conservatism is appropriate (21).

The final step in Poston’s account of conservatism is to identify which, if any, actual cases of belief are cases of empty symmetrical evidence. Poston suggests that the sort of “hinge propositions” that Wittgenstein discusses in On Certainty fit the bill.1 These are propositions that we do not inquire into because they are always already assumed within our practices of inquiry. Examples include “Memory is reliable,” “Valid inference is reliable,” “The meanings of terms are generally stable,” and “The future will resemble the past.” Poston’s strategy is to use propositions such as these—and not the deliverances of the senses—to ground his epistemological system (21–23).

3. Two Alternate Systems

In this section, I accept Poston’s arguments for conservatism, but argue that he underestimates the number of hinge propositions that can be believed in cases of empty symmetrical evidence. In particular, by looking at William James’s “The Will to Believe” I suggest that there is a set of hinge propositions that is distinct from the one that Poston highlights. I argue that realizing this leads to a version of the alternate systems objection to Poston’s coherentism: although he has developed a coherent epistemological system, Poston has not given us reason to suppose that he has developed the true epistemological system.

In “The Will to Believe,” James starts at much the same place that Poston does. James considers the question of how evidence relates to justified belief, and draws a number of distinctions between relevant cases of believing. James chooses to focus on cases where there is what he calls a genuine option between different belief contents—that is, cases where the contents under consideration are all plausible to the believer, the range of contents under consideration is exhaustive, and the consequences of believing at least one of the contents is significant. The genuine option that James has most clearly in mind is that between having religious faith and lacking it. Both alternatives are plausible to James, there is no third alternative because choosing not to choose between them is to lack faith, and both would have marked effects on how James lives his life.

What is most important for our purposes is when James remarks that “there are, then, cases where a fact cannot come at all unless a preliminary faith exists in its coming.”2 One plausible understanding of this claim is that although faith serves as a precondition for such facts coming to be, faith itself must be held on empty symmetrical evidence. Before one has faith, there is no evidence because the facts that support faith have not come into existence, and to take such facts as evidence after one has faith would be viciously circular. Therefore, the central tenants of one’s faith would constitute hinge propositions of the sort that Poston discusses.

However, when we look at specific faithful hinge propositions, we see that their content is radically different from those that Poston acknowledges. Whereas the propositions that Poston mentions are those that make possible induction and inference to the best explanation, the faithful hinge propositions that James has in mind would be ones such as “Everything has a purpose,” “Things act intelligently,” or “There is a rational order to things”—that is, propositions that make possible teleological and intentional explanations.3These different explanatory strategies lead to vastly different explanations of things.

The easiest way of understanding this state of affairs is that there are two alternate systems at play here. One is the system based on induction and inference to the best explanation that Poston defends, and the other would be a Jamesian system based on teleological and intentional explanations. Given Poston’s sympathy towards Quine’s rejection of the analytic/synthetic distinction, it would appear that he must concede that this Jamesian system can be made internally coherent through adjustments in the overall “web of belief.” (85-87) Therefore, we have two alternate systems that are both coherent, and the question is which one is the true system.

Poston’s obvious first response would be to argue that the system he defends better meets his preferred explanatory virtues—those of conservatism, simplicity, and explanatory power (80–85)—but try as I might I cannot see a non-question-begging way of understanding any of these virtues that would settle the issue. Conceivably, Poston and a Jamesian can each give an equally circular argument for the virtues of their systems, and we are back to the question of truth.

4. The Trilemma

At this point, Poston finds himself in a precarious position. He has developed his system in isolation from the Jamesian system, so it is understandable that he has previously believed the hinge propositions of his system on empty symmetrical evidence. However, now that he has become aware of the Jamesian system, the evidential situation appears to have changed. The existence of an alternate system calls into question the truth of both systems’ hinge propositions, and this alternate systems objection becomes a piece of evidence that is equally weighted against both systems. Therefore, the appearance is that Poston now believes his hinge propositions on symmetrical evidence that is not empty. This leaves Poston with three options, all of which come with a cost.

The first option is for Poston to reject this appearance of the evidential situation without providing a reason in support of his rejection. The benefit of doing this is that Poston can maintain that he believes his hinge propositions on empty symmetrical evidence, and therefore that his beliefs are conservatively justified. The downside is that this is a paradigm case of dogmatism in its pejorative sense, as this amounts to ignoring new evidence for no reason other than that it calls one’s beliefs into question.

The second option is for Poston to acknowledge the appearance of the evidential situation, and to offer some reason either as to why the appearance is deceptive or as to how his set of hinge propositions is superior. The benefit of doing this is that Poston can maintain the justification for his beliefs without being intellectually vicious. But there are two downsides. First, such reasons are few and far between, leaving Poston the unenviable task of identifying them. Second, and more pressing, even if Poston does identify these reasons, in doing so he would convert the evidential situation to one of asymmetrical evidence, thus leaving his conservatism as an idle wheel within his system. The lesson of pursuing this option would seem to be that conservatively justified propositions need not play a role in Poston’s defense of coherentism, but his response to the Basic Reasons Dilemma entails that they must play an essential role.

The last option is also the most obvious: Poston can merely accept the appearance without attempting to change the evidential situation one way or the other. To my mind, this is the most promising option for Poston, but it too is not without cost. It is promising because it avoids the downsides of the first two options—by accepting the appearance Poston would not be intellectually vicious and by refraining from altering the evidential situation Poston does not engage in the sort of arguing that is counterproductive to his conservatism. However, the cost is a contingent or contextual form of skepticism. So long as Poston is unacquainted with any alternate systems, he can justifiedly believe his hinge propositions thanks to his conservatism. However, because suspension of belief is the proper response to cases of non-empty symmetrical evidence, once he becomes acquainted with such systems his beliefs will lose their justification.

To summarize my line of critique, I concede that Poston solves the Basic Reasons Dilemma if his conservatism is well founded. However, his conservatism depends on the relevant evidential situation being one of empty symmetrical evidence, and the alternate systems objection transforms the situation into one of non-empty symmetrical evidence. At this point, Poston faces a trilemma, the horns of which are dogmatism, self-defeat, and skepticism. None are promising options for Poston’s conservatism, and the worry is that without his conservatism Poston no longer has a sufficient response to the Basic Reasons Dilemma.


James, William. The Will to Believe and Other Essays in Popular Philosophy. Cambridge: Harvard University Press, 1979.

Poston, Ted. Reason and Explanation: A Defense of Explanatory Coherentism. Basingstoke, Hampshire, UK: Palgrave Macmillan, 2014.

Wittgenstein, Ludwig. On Certainty. Edited by G. E. M. Anscombe and G. H. von Wright. Translated by Denis Paul and G. E. M. Anscombe. New York: Harper Torchbooks, 1972.

  1. Wittgenstein, On Certainty, §152 and §341–43.

  2. James, “Will to Believe,” 29.

  3. See James’s distinction between It and Thou at James, “Will to Believe,” 31.

  • Avatar

    Ted Poston


    Rival Explanatory Paradigms and Justification: A Response to Dabay

    Thomas Dabay provides a thoughtful and interesting perspective on my explanationist view. He focuses on the alternative systems objection to coherentism and argues that this is particularly problematic given my views about epistemic conservatism. Traditionally, the alternative systems objection targets coherentist views of justification because typical coherentist views hold that the justification of any belief is entirely a matter of its internal relations to other beliefs. The objection continues by observing that lots of different sets of beliefs—like a good work of fiction—bear virtuous internal relations to each other member of the set. But, presumably, to be epistemically justified in a belief requires more than its being embedded in a coherent work of fiction. Epistemic justification requires more than internal relations among beliefs in one’s doxastic system.

    I briefly discuss the alternative systems objection in my book (10–12). I argue that the objection is not especially problematic for a non-doxastic coherentist view according to which coherence must be assessed over beliefs and experiences. Experiences stand in a causal connection to the external world (beliefs do as well, for that matter), and so a coherentist is not committed to the claim that the justification of a belief is analogous to a proposition in a good work of fiction.

    Any attempt to revive the alternative systems objection against coherentism is bound to fail. What made the objection persuasive was the idea that a good work of fiction can bear all the epistemically relevant relations a coherentist claims matter for justification. But non-doxastic coherentism claims the coherence must be assessed over all assertive propositional contents. Fiction is not coherent with experience (nor belief either). Thus revised, the alternative systems objection becomes the jejune observation that coherentism implies that radically alternative systems may be epistemically justified. But that’s an objection to coherentism only if the truth counts against a view.

    The strength of Dabay’s comments lays in his interesting remarks centered on William James’s remarks on faith and my attempt to explicate epistemic conservatism by way of the notion of empty evidence. James thought that we were permitted to believe certain claims in cases in which a belief was a live, forced, and momentous option. This is analogous to my claim that in the case of empty evidence we are permitted to continue to believe a claim even though it is unsupported by positive evidence. James thought that belief in some cases was permitted because otherwise certain goods could not be achieved. If one is in ill health and will recover only if one believes that one will recover then James held that belief is permissible. James may have had in mind a conception of permissibility that included practical considerations. In my remarks on empty evidence, I focus on an epistemic permission to continue to believe. But it does seem that both James and I recognize that there is a way to positively evaluate belief apart from direct evidential considerations.

    Another agreement with James’s remarks on faith is that some beliefs held in the state of empty evidence are momentous. In my case, an apt analogy is the way a subjective Bayesian thinks about evidence. Such a Bayesian holds that evidence is probabilistic relevance: e is evidence for p just in case Pr(p|e) ≠ Pr(p). Evidence then requires probabilities. But where do the probabilities comes from? Subjective Bayesians claim they come from belief. One’s initial probability function is not based on evidence; rather it is precondition for evidence. On my view, one’s initial beliefs are not based on evidence; rather they are preconditions for evidentially based beliefs. One’s initial beliefs aim for explanatory integration. If they cannot achieve integration then they should be abandoned. We see here the teleological role of belief.

    Dabay also highlights a problem with my remarks about empty evidence and alternative hinge propositions. The general problem is this: if there are alternative hinge propositions and such propositions are held in the state of empty evidence then there are no evidential differences between such propositions. Thus, my view implies that the resulting systems of belief cannot be compared on their evidential grounds. Dabay argues that my view can avoid this problematic response only if I give up on the idea that alternative hinge propositions are held in empty evidence.

    The problem that Dabay latches onto is the implications of my explanationist view concerning the problem of pluralism. We know that there are significant differences in the opinions of reasonable people. There are differences in fundamental explanatory projects and, perhaps, some of these differences are due to alternative hinge propositions. My view implies that each person has a right to maintain alternative hinge propositions in the state of empty evidence. In such a state, there is no evidential reason for preferring one hinge over another. Yet belief aims for explanatory integration and so each person is under an obligation to inquire further. In the state of empty evidence it remains to be seen whether the hinge propositions can bear the weight of explanation. Thus, my view doesn’t imply that there can be no epistemic reasons to prefer one view over another. Rational conversion is possible because one can come to see that one’s initial view isn’t explanatorily virtuous.

    Another Bayesian analogy is instructive. Given an initial probability function, it’s natural to think that one should update one’s probability function in light of new evidence in a way that respects one’s current opinions. Thus, one should follow the diachronic form of Bayes’s theorem. Given some new evidence E relevant to a proposition p, one’s new probability for p—Prnew(p)—should equal one’s old probability for p given e—Prold(p|e). While this is a natural thought, sometimes new evidence can lead you to think that your initial opinions were mistaken. In such a case one may think that it’d be crazy to follow your former opinions. The phenomenon here can be captured by the way we can trace out implications of some beliefs. We often attempt to reason people out of certain beliefs by observing that “if you really thought that then you should think this other thing as well, but that would be crazy.” The point here is that there can be legitimate ways of criticizing alternative hinge propositions by tracing out the implications of such views.

    The analogy shows us then that there are rational ways of criticizing rival explanatory paradigms without being committed to foundationalism. An explanatory coherentist may hold that some rivals paradigms are not as virtuous as others. The holistic properties of rival paradigms enables rational differences between them.



Disagreement and Epistemological Guidance

1. Introduction

Epistemic conservativism, the view that belief in certain situations can be sufficient for justification, has faced plenty of criticism. The version defended by Ted Poston (2012 and 2014), however, which I will call mere-belief conservativism (MBC), ably avoids several of the standard arguments that have been directed at the view. My goal in this paper is to add myself to the list of critics by arguing that even Poston’s MBC is problematic. I do this in three stages. In section 2 I argue that Poston’s reply to several common worries about conservativism makes MBC vulnerable to an ad hoc objection. In section 3 I argue that Poston’s reply to a warranted-assertion objection is unsuccessful. Then in section 4 I argue that the considerations that Poston marshals in his positive case for MBC are not persuasive. My conclusion is that mere-belief conservativism is no more appealing than conservativism in its other forms.

2. The Ad Hoc Objection

Poston’s version of conservativism, MBC, holds that belief is sufficient for justification just in cases of “mere belief,” that is, in cases of what Poston calls empty symmetrical evidence (Poston 2014: 21).1Those are cases in which a subject has no evidence (no non-circular evidence) for or against the proposition. There are not many cases of beliefs in empty symmetrical evidence; MBC is thus a highly restricted view. Poston references Wittgensteinian “hinge propositions” as examples, such as the belief that memory or induction is reliable, beliefs for which any evidence is bound to be question begging (21–22).2

Since MBC is highly restricted, it avoids many counterexamples to conservatism in general. For example, Christensen (2000) presents the following sort of case: a scientist believes that one theory is correct because it fits the available evidence, but then she realizes that a competing theory equally fits the evidence. Since they fit the evidence identically, one should not believe one of these theories over the other; but one is justified in doing so according to conservativism, because one’s belief in the original theory justifies one in retaining that belief (or assigning it a higher credence).

As Poston points out, this counterexample is no threat to MBC, since the example is not one of empty symmetrical evidence; rather, it is a case of positive symmetrical evidence (31–32). Such a defense, which Poston uses repeatedly, raises an important question: what exactly is Poston’s justification for restricting conservativism to mere beliefs? That it avoids the counterexamples, of course, is one possible justification. Philosophy often proceeds that way. The danger, though, is twofold. First, such ad hoc restrictions avoid counterexamples of the past, not necessarily counterexamples of the future. One could push this objection by adhering to Poston’s constraints on mere belief: in any case of empty symmetrical evidence for p, there will also be empty symmetrical evidence for not-p. Mere belief in not-p would then be just as justified as mere belief in p (ceteris paribus).

The second danger, however, is more important: the resulting ad hoc view might be inconsistent with its motivations in the first place. Here are four considerations that make this danger pressing. First, Poston thinks that we can see how MBC works via the case (37–38; originally from Harman 1986) of a subject, Karen, who acquires a belief that leads to the “discovery” of all sorts of explanatory connections in her belief set (those explanatory connections are justification-conferring conditions on Poston’s general explanatory coherentist view). Later the subject learns of evidence that undermines the original belief. Poston holds, though, that she is justified in retaining that belief. Now, we might not find this example compelling in the first place, but the key point is that Poston cannot use it to explicate or motivate his position, since his position depends on ad hoc restrictions that are precisely those that make it incompatible with the example, since the example is not a case of empty symmetrical evidence.

The second consideration concerns Poston’s remarks that conservativism is a “principle of last resort,” an idea which he takes from Sklar (1975). The idea is that we are allowed to appeal to belief as a justifying factor if evidential considerations “fail to motivate a decision for us” (27; Sklar 1975: 375). So stated, this principle seems to apply to cases of positive symmetrical evidence just as much as empty symmetrical evidence. Again, then, the restrictions to empty symmetrical evidence seem ad hoc.

Elsewhere in the text Poston uses the principle of last resort in an MBC-specific way, and this use leads to the third consideration that raises ad hoc worries. In order to restrict the principle of last resort to MBC, Poston holds that whenever “evidential considerations are relevant” one would not be justified in holding one’s belief merely conservatively but instead should withhold judgment (32). Thus the mere relevance of evidential considerations “swamps” and “destroys” conservative justification. Of course, it is natural to imagine that evidential considerations are always relevant to belief, which by Poston’s principle would suggest the natural conclusion that there is no such thing as conservative justification. It is therefore essential for Poston to distance himself from evidentialism (roughly, the view that justification is determined by evidence), because if evidentialism is true then conservativism, even MBC, is false. Unfortunately, when Poston discusses evidentialism (specifically, Conee and Feldman’s (2004) position, with whom Poston shares a mentalist conception of evidence), he claims that the reason why evidentialism is false is merely that it conflicts with MBC (94–95). He states:

In the special case of empty symmetrical evidence E [evidentialism] yields the wrong verdict; in all other cases E is a sound principle. (95)

It therefore isn’t true, as Poston claims, that Poston’s view inherits the motivations and plausibility of evidentialist accounts of justification (see ch. 4), because the motivations and plausibility of evidentialism aren’t compatible with MBC.

Before moving on, I should answer a reply Poston would make here. Poston claims that his view is compatible with Conee and Feldman’s strong supervenience principle (roughly, that justification strongly supervenes on one’s evidence), and therefore even though he doesn’t capture their evidentialist thesis he does capture a key part of their view (94). Poston’s key idea here is that justification can supervene on evidence without justification always determined by (in the sense of requiring) evidence, since he thinks that conservative justification is non-evidential. It is no doubt true that that supervenience claim is a key part of Conee and Feldman’s view. Poston quotes their observation that it is “bedrock evidentialism” (95; Conee and Feldman 2004: 102). But there is a complication here: Conee and Feldman say that the supervenience claim is bedrock because it is difficult to formulate their specific evidentialist commitment precisely, not because they think that there could be a case, as Poston does, of a belief that isn’t justified by evidence (see Conee and Feldman 2004: 101–2). So it just doesn’t follow that Poston has captured the core motivations for evidentialism. Indeed, this issue arises later in the book, where Poston himself relies on evidentialist motivations that reveal again the ad hoc nature of MBC. For example, Poston uses BonJour’s (1985) anti-foundationalist argument for coherentism, which plays a critical role in Poston’s case for his explanatory coherentist view (ch. 5). A premise of that argument that BonJour and Conee and Feldman endorse is that justification requires good reasons or evidence (111). That is exactly the idea that Poston cannot consistently rely on.

Now I turn to the fourth and final consideration that suggests that the ad hoc nature of MBC is inconsistent with its motivations. This fourth consideration comes from one of Poston’s positive arguments for MBC. The argument is that “our epistemic practices are significantly conservative” (38–39). Poston continues:

When we take into account the force of new evidence we follow the maxim of minimal mutilation. This maxim is firmly entrenched in our epistemic practices. When subjects affirm the consequent, they can learn that this form of reasoning is faulty. Further, they can learn to replace affirming the consequent with explanatory reasoning which explicitly considers relevant alternatives to one’s favored conclusion. Yet conservativism differs from natural reasoning errors. Subjects are always and everywhere relying on their antecedent convictions. Even the effort to improve reasoning relies on our trust in unchallenged forms of reasoning. (39)

These remarks provide no support to MBC. MBC is based on ad hoc restrictions of conservative justification to mere beliefs, beliefs for which one fails to have (and perhaps it is not possible to have) non-circular evidence. The trouble is that the sort of general epistemic practices at issue here, and the fact that improving one’s reasoning relies on trust in some other form of reasoning, have nothing to do with MBC’s ad hoc restrictions to empty symmetrical evidence. One can have all sorts of evidence regarding different forms of reasoning, and Poston explicitly discusses in the passage a process of accounting for evidence. Poston simply isn’t adverting to cases of mere belief and empty symmetrical evidence.

In this section I have argued that Poston’s defense of MBC depends essentially on certain ad hoc restrictions, yet, as this objection contends, those restrictions are not well motivated and indeed are in tension with (i) the examples he uses to develop and motivate the view, (ii) the principle of last resort, and (iii) his positive arguments for the position.

3. The Warranted-Assertability Objection

A different problem for MBC concerns Poston’s response to the warranted-assertability objection of Christensen (1994). Christensen holds that conservativism has the problematic consequence that one can correctly say, “I happen to believe it—and that is part of my justification for continuing to believe it.” Poston’s strategy is to argue that MBC does not have this consequence (28). He does so by arguing that Christensen’s claim is, for Gricean reasons, not appropriately assertable (29).

There is an important unclarity in Poston’s reply, however. Incorporating Gricean considerations forces us to recognize two senses of “correctly” saying something. It can be correct in the sense of true, or correct in the sense of conversationally appropriate. A charitable reading of Christensen’s claim is that one can truly say, “I happen to believe it—and that is part of my justification for continuing to believe it,” and Poston never shows that that is wrong. So in a sense Poston hasn’t answered Christensen’s worry. When Poston claims that in certain contexts “it is wrong to say that one is justified in believing p, even though one can remain justified in believing p” (29), Poston is using “wrong” in the sense of inappropriate (not in the sense of not true).

Poston thinks he can parry Christensen’s worry by holding that “when one has merely conservative justification for p, one’s conservative justification is unasssertible” (29). Note that Poston cannot be claiming there that any belief with conservative justification is unasssertible: he has been asserting all book that we have specific beliefs with conservative justification! So it is key to his claim above that it is just beliefs with “mere” conservative justification that are unasssertible (since all conservative justification for him is mere-belief conservative justification, there is a double “mere” here). The implausibility of this claim is rather over determined. First, appeals to Gricean considerations are unpersuasive if they are ad hoc, which is why Poston holds that his appeal is based on the conversational norm to be relevant. Poston claims that one cannot cite one’s belief and be relevant: when p is at issue, one’s belief that p is “entirely irrelevant” to whether p is true (29). That is a fascinating claim, since it raises the obvious question of why the belief can justify the belief that p if the belief is entirely irrelevant to whether it is true. The consequence is that justification does not retain a connection to truth, which is in tension with a basic aspect of justification shared by internalists and externalists and Poston himself. Poston can claim that his explanationist coherentist account can in some cases maintain a connection between justification and truth (in some sense or other of “connection” and “truth”), but it is an inescapable fact then that the whole coherentist edifice depends on a fundamental level on justification that has nothing to do with truth.

Second, Poston notes that a fact about Gricean implicatures is that they are cancellable, and he claims that the purported implicatures generated by assertions of conservative justification have this feature. Poston uses an example of a mathematical claim, like Goldbach’s conjecture, which one might believe with justification even though one cannot prove it. One might say, according to Poston, “I maintain this belief even though I don’t have any special evidence for it. I find myself believing that Goldbach’s conjecture is true” (30). Again we are in a position, however, in which Poston’s example does not establish what it is supposed to. For one thing, the mathematician making this claim does have evidence that Goldbach’s conjecture is true. The conjecture has been computer tested on numbers over 1018. Could that evidence prove the mathematical theorem? No. But can that evidence still convince many mathematicians? Yes. Furthermore, mathematicians have track-record evidence for certain types of (perhaps inductively or intuitively supported) beliefs that have then been proven—Fermat’s last theorem, which Poston himself invokes, is a good example of such. What Poston really needs is a totally different sort of example, which, rather than helping MBC, shows why MBC is problematic. Imagine a mathematician who has no evidence at all for or against believing a certain theorem—not even intuitive or inductive grounds, nothing—but finds herself believing it anyway. Then Poston is committed to the implicature-canceling assertion she can make: “I believe this theorem, and I am justified in doing so just because I believe it, even though I have no evidence at all for thinking that it is true.” Few epistemologists will find such a position acceptable.

Ironically, we can now see that Christensen’s objection to conservativism isn’t quite an objection to MBC, but a reformulated warranted-assertability objection does apply to MBC and is even stronger. On Poston’s account, assertions conservative justification that violate relevance will be cases in which the conservative justification isn’t just a part of one’s justification—it will be the entire justification. So Christensen’s worry isn’t quite right, in claiming that it is true to assert, “I happen to believe it—and that is part of my justification for continuing to believe it” (my italics). Instead, Poston is committed to the even more problematic implicature-cancelling true assertion: “I believe this theorem and my belief is my entire justification for continuing to believe it, even though I have absolutely no evidence for it.”

In this section I have argued that Poston’s reply to the warranted-assertability objection fails. Far from clarifying why MBC is unproblematic, it actually helps make clear why it is problematic.

4. The Positive Case for MBC

Poston presents a positive case for MBC based on three considerations. The first, based on our epistemic practices, I already discussed above; in fact, it is not an argument for MBC at all. The second is based on avoiding skepticism. The argument here depends on establishing (i) that MBC actually can avoid skepticism and (ii) that no other theory can. Given Poston’s defense of MBC, the dialectical situation here is quite problematic. The conservative principle of last resort is essentially a stipulation that certain beliefs are conservatively justified because, given certain assumptions, there is no other way to understand them as justified. Many views, though, can stipulate skepticism away dogmatically. So any claim to dialectical advantage here is hard to appreciate.

I will focus my attention, then, on the final consideration that Poston marshals to support MBC: Poston contends that the epistemic internalist idea of the subject’s perspective shows that MBC is true. It won’t surprise the reader to hear that I don’t find this argument compelling. I will argue that even if we grant the perspectival nature of justification that Poston is interested in, it provides no support at all for MBC.

First, let’s examine the various points Poston makes about internalism and the subject’s perspective. Poston characterizes it this way:

The perspectival character of justification is that the epistemic justification of a subject’s beliefs depends on a subject’s perspective. Facts that a subject is unaware of do not make a difference to the justificatory status of a subject’s beliefs since they are not part of her perspective. (39)

I will grant the idea here: there must be grounds (or facts) that the subject is aware of from her perspective in order for her to have a justified belief. Many internalists hold as much. Prima facie, however, this understanding of justification supports the idea that justification requires evidence or reasons (as noted by BonJour, Conee, and Feldman). So on the surface the connection between justification and the subject’s perspective invoked by Poston actually suggests that MBC is false, not true (since conservative justification is non-evidential). It is therefore important to examine carefully how Poston thinks such a notion supports his conservativism. Poston’s text continues:

The facts relevant for determining justification are facts about the subject’s perspective. (39)

From the first move in this argument we are in trouble. Contrary to Poston, this claim is not the same as, and doesn’t even follow from, the previous claim. Just because justification depends on the subject’s perspective, and subjects must be aware of some fact in order for it to contribute to her justification, it does not follow that the facts at issue are about the subject. I might need access to the fact that the streets are wet in order for it to be my reason for believing that it rained. It does not follow that my grounds for believing that it rained is a fact about me, rather than a fact about the streets.

Next, Poston claims that in order for experience to justify one’s beliefs, one must believe that experience is reliable or veridical (39–40; Poston references claims of Goldman [1979] and Lehrer [1974]). But then Poston claims:

The point of both Goldman’s and Lehrer’s remarks is that justification is perspectival. We rely on our beliefs about what counts as good sources of evidence to determine what beliefs are epistemically good to hold. (40)

Unfortunately, it is not the point of Goldman’s and Lehrer’s claims that justification is perspectival in the sense that Poston has just specified. The idea that subjects rely on (or, as Poston also says, require) second-order beliefs about reliability or what is good reason to believe what is not at all the same idea as what Poston identified before as the perspectival character of justification (that a consideration cannot justify a subject’s belief unless she is aware of it).

Next, Poston holds that the distinguishing characteristic of epistemic justification is that epistemic justification is tied to the standard of truth, and furthermore that the relevant notion of truth must rely on a notion of right reason (40–41). This claim is surprising in two ways. First, it is surprising because it is another distinct claim from what Poston has said before, and does not, contrary to Poston’s claims, provide the “normative backing” for the points made before. Second, it is in tension with basic facts about MBC. The idea Poston references in BonJour (1985: 5–6) here is BonJour’s idea that epistemic justification is the notion of belief based on truth-conducive reasons. That is exactly what MBC does not provide.3 When Poston then infers that MBC is the lesson we should draw from the perspectival nature of justification, it is quite mysterious to the reader what is going on or where at inference even comes from.

There are two distinct problems here. One is that Poston invokes at least four different epistemological ideas and moves between them in problematic ways. It does violence to the claims made by Goldman, Lehrer, and BonJour to assimilate or conflate them as Poston does (nor does Poston’s commentary illuminate the connections). Second, in no way do any of these remarks actually show that MBC it true. In fact, some of them even suggest the opposite. Poston’s positive case for MBC is therefore not compelling, and does not provide the reader with any grounds for thinking that it is true.

5. Conclusion

I have argued that Poston’s case for mere-belief conservativism fails. The view faces a variety of objections, including ad hoc and warranted-assertability objections, and the positive arguments for it are unsuccessful.


BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge: Harvard University Press.

Christensen, D. 1994. “Conservativism in Epistemology.” Nous 28: 69–89.

Christensen, D. 2000. “Diachronic Coherence versus Epistemic Impartiality.” Philosophical Review 109: 349–371.

Conee, E., and R. Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford University Press.

Goldman, A. 1979. “What Is Justified Belief.” In Justification and Knowledge, G. Harman, 1986. Change in View: Principles of Reasoning. Cambridge: MIT Press.

Lehrer, K. 1974. Knowledge. Oxford: Oxford University Press.

Poston, T. 2012. “Is There an ‘I’ in Epistemology?” Dialectica 66: 517–541.

Poston, T. 2014. Reason and Explanation: A Defense of Explanatory Coherentism. New York: Palgrave Macmillan.

Sklar, L. 1975. “Methodological Conservativism.” Philosophical Review 84: 374–400.

  1. All further references to Poston are to this work and will be given as page numbers only.

  2. Interestingly, it isn’t so clear that any evidence for every such hinge proposition must be circular. One can imagine a view on which the trustworthiness of one’s senses legitimizes testimonial evidence, and via testimony one gets evidence for memory and induction. One is still pivoting the edifice of belief on the original hinge of the trustworthiness of the senses, however, so Poston would likely be undeterred by this possibility.

  3. Poston could try to invoke a distinction between evidence and reasons here, but that would be a misunderstanding of the position of BonJour, as well as that of Conee and Feldman.

  • Avatar

    Ted Poston


    Making Conservatism Great Again: A Reply to Schnee


    I appreciate Ian Schnee’s forceful criticisms of my attempt to explicate a plausible version of epistemic conservatism. As each commentator has pointed out, epistemic conservatism plays a pivotal role for my coherentist theory and so deserves careful attention. I argue that a belief’s justification is a matter of its fit in an explanatory coherent system that beats relevant competitors. Moreover, I argue that a belief’s justification is always relative to a set of background beliefs. I contend that unless background beliefs have some level of justification simply in virtue of being held then skepticism follows. The key is to formulate a plausible version of conservatism that does not do violence to our firm judgments about the role of evidence in justification. I’ve argued that the core conservative claim is a coherence condition on a subject’s mental life. Unless a subject has a special reason to change her views, she has a right to continue to maintain those views. Or, as I put it, if a subject believes p in the state of empty evidence she has a right to continue to believe p. This epistemic right is not indefeasible. As I mentioned in my responses to other commentators, belief is teleologically ordered to knowledge. Consequently, such a coherence condition on a subject’s mental states is not the be and end all to epistemology. I honestly think epistemic conservatism, properly understood, makes good epistemology. Schnee disagrees. Let us therefore reason about our differences.

    Schnee’s first charge against my version of conservatism is that the restriction to empty evidence is ad hoc. As I see it, respect for the evidence requires that when one has evidence one’s attitudes should track the evidence. However, there are cases in which one lacks evidence entirely. Think of an initial choice of a probability function or an initial choice of natural kind terms. In such cases, I argue that one has a right to one’s initial positions. One has a right to see if those views will bear the weight of explanation. Schnee finds this restriction ah hoc. It is unclear what to make of the charge of ad hocery. Being ad hoc is compatible with being true. And because I do offer reasons for the restriction, one may well think that either the restriction isn’t ad hoc or that being ad hoc is even compatible with being theoretically motivated.

    Schnee’s charge of ad hocery does morph into a more specific criticism: that the empty evidence restriction is inconsistent with its motivations. He points to four considerations to develop the charge of inconsistency. He first picks up on my discussion of Harman’s case of Karen (see 37–38). Schnee observes that this case does not motivate the empty evidence restriction because in Harman’s case Karen does have positive evidence. Of course, in my book I explicitly say that Harman’s case does not motivate the restriction for precisely the reason that it is not a case of empty evidence (38). My purpose in mentioning Harman’s case is to look at the way belief plays a role in explanatory coherence. Second, Schnee observes that my mention of Sklar’s principle of last resort—“belief is a justifying factor when evidential principles fail to motivate a decision”—doesn’t uniquely motivate the empty evidence constraint. I don’t rely on Sklar’s principle to motivate the restriction. The dynamic I see at play in Sklar’s principle is the implicit recognition that epistemic conservatism is a unique epistemological principle; it is not an evidential principle, but it is a principle of a different character. On my view, it is a coherence constraint on doxastic attitudes. Third, Schnee picks up on the tension between conservatism and evidentialism. Schnee’s remarks are similar to Andrew Cling’s and so I refer the reader to my discussion of Cling. The fourth and final consideration that Schnee offers centers on epistemic practice. Roughly, I argue, following Lehrer, that there is no exit from the circle of belief. There is no epistemic bedrock. We always rely on background beliefs. Schnee says that this doesn’t motivate empty evidence conservatism. That’s a different charge from the charge of inconsistency. As to the different issue of motivation, it is plausible that if there is no exit from the circle of belief and there are epistemic dependence relations between beliefs, then some beliefs will be held in a condition of empty evidence. This isn’t a formal argument—it’s much too quick for that—but it is a plausibility consideration. I refer to the reader to the Bayesian analogy on evidential relevance and the role of belief I offer in my response to Dabay.

    The second criticism Schnee develops is that my response to Christensen’s objection to conservatism fails. Christensen argues that conservatism implies that one can correctly say, “I believe p and that’s my justification for it.” In my book, I respond in Gricean fashion. There are two ways assertions can be said to be correct. They can be true or they can be conversationally appropriate (inclusive “or”). I argue that Christensen’s assertion is not conversationally appropriate because when one has only conservative justification one has nothing informative to offer one’s interlocutor. One’s interlocutor already knows that you believe p and so Christensen’s assertion does not move the conversation along. Schnee observes that Christensen’s assertion may still be true (though inappropriate) and so we can take Christensen as insisting that such an assertion is never true. The problem with this suggestion, though, is it doesn’t advance the argument against conservatism. My appeal to Gricean considerations highlights that the implausibility of the Christensen assertion may be responsive to inappropriateness, not to falsity. This is an undermining move, not a rebutting move. Of course, one can insist that such an assertion is always false, but then it looks like the objection just turns into an insistence that conservatism is false.

    Schnee also resists my attempt to show a way in which one can cancel any Gricean implication that afflicts Christensen’s assertion. I consider a case in which a mathematician believes Goldbach’s conjecture but has no grounds for it (and no grounds against it). Schnee observes that mathematicians have lots of evidence that the conjecture is true. First, this involves a controversial position on what counts as evidence for necessary truths. As Aristotle says, “It is foolish to accept probable reasoning from a mathematician” (Nicomachean Ethics, Book I, chapter 3). Second, the kind of case I was thinking about is much more homey. Mathematicians have beliefs about theorems prior to anything even approximating inductive evidence coming in. The theorem just seems natural or right; it is believed. In such a case it’s natural to say, “I believe it, though I don’t have any reason for it; it just seems right.” One might think that what is really going on here is that the mathematician has some kind of special experience—an intuition or a seeming. And so there’s really positive evidence. I resist this move and argue that the relevant mental state is belief, but that is a story for another occasion.

    The last criticism Schnee develops is that my attempt to provide a positive argument for conservatism from the perspectival character of justification fails. It is unclear what argument Schnee advances beyond the observation that there are other internalist views that resist epistemic conservatism. Schnee mentions Bonjour’s internalism and Feldman and Conee’s internalism, both of which attempt to uphold the idea that the facts that determine justification are perspectival without adopting conservatism. I do argue extensively in chapter 5, “Bonjour and the Myth of the Given,” against such internalist views. I refer the reader to that chapter for relevant arguments against such internalist views. If internalism is true and there is no epistemic bedrock then, I contend, an explanatory coherentist view is motivated and plausible.

    Let me close with two observations about the empty evidence restriction. First, Schnee does not mention my engagement with Christensen’s core objection to epistemic conservatism, viz., that there’s no “I” in epistemology (see 30–34). Christensen argues that the identity of the subject is irrelevant to epistemology. As I mention in the book, I think Christensen has shed light on perhaps the central worry with epistemic conservatism. If his worry is removed, then we have removed the thorn in the side of a conservative epistemology. Second, anyone broadly moved by a Cartesian epistemology wants justification to track reasons that are found in a subject’s perspective. The key question is whether we always have reasons for every belief. I’ve argued that we do not. If those arguments are correct then a broadly Cartesian epistemology must find another natural element of a subject’s perspective that provides some justification. I’ve argued that, suitably understood, belief fits that role. Inquiry starts with belief but ends with knowledge. If the view I develop in Reason and Explanation is correct then the kind of justification required for knowledge is virtuous explanatory coherence.

    • Avatar

      Jonathan D. Matheson


      Reply to Nathan Ballantyne

      Ballantyne has concerns regarding whether the principles advocated in the book can help guide individuals in the real world. Ballantyne claims,

      Many popular epistemological theories would be successful, on their own terms, by accurately describing facts about the nature of knowledge, epistemic justification, evidence, and the like. Epistemologists who articulate and defend such theories are not trying to “do” anything with their theories. Their aim is to describe something, period. But Matheson apparently practices a different brand of epistemology. He is trying to improve intellectual life by describing correct epistemic principles.

      It is important to distinguish two projects: (i) uncovering true epistemic principles, and (ii) building better thinkers. One might hope to do the latter by way of the former, but as Ballantyne cautions, we have reason to be skeptical here. Ballantyne maintains that approving of principles (epistemic or otherwise) is insufficient to be guided by them. I agree. I confess that I have not abandoned my deeply controversial political, religious, scientific, or philosophical beliefs. Despite being convinced by my book, those beliefs remain.

      Perhaps this comes as a disappointment, but Ballantyne has described the project as much more ambitious than it is. Rather than primarily being about building better thinkers, the goal of the book is simply to correctly describe some epistemic facts. Those facts are about the nature and strength of a particular defeater—a defeater one gains when one learns of a disagreement. The arguments concern the existence of such a defeater, its strength, and the conditions under which it is itself defeated. I argue that this defeater has consequences for which of our beliefs are epistemically justified. More carefully, I argue that this defeater has consequences for which of our beliefs enjoy synchronic epistemic justification, where I argue that this is but one kind of epistemic justification. While I do argue for and endorse various epistemic principles along the way, I also note that this is an incomplete picture. Plausibly there are a number of principles that are relevant to what we should believe, some moral, some pragmatic, and even a variety of epistemic principles. The book’s focus is on one kind of epistemic principle—the one relating to synchronic epistemic justification. The project cannot be straightforwardly action guiding since it is silent on what other principles may govern belief and how all these principles interact. Do some outweigh others? Do some altogether trump others? The book is silent on all of these matters. Only when coupled with answers to questions such as these, can the principles endorsed in the book be part of a package that delivers all-things-considered verdicts about what one should believe. Such a consequence is simply stronger than the arguments warrant.

      In these ways I take it that my project falls right in line with a great deal of contemporary epistemology. Evidentialists argue that epistemically justified beliefs are those that fit the subject’s evidence. Reliabilists disagree and argue that it is only the belief forming processes that are relevant. I take it that neither are under the impression that their arguments will make readers better align their beliefs with their evidence or adopt more reliable belief forming processes, even though evidentialists and reliabilists may motivate their theories by looking at idealized fictional cases and apply their theories to the real world. My project is neutral as to what features generate epistemic justification. My main goal is to argue that we have good reason to believe that there is a powerful defeater for many of our cherished beliefs. These arguments may not alter how people think, but I hope to have described another aspect of epistemological reality.

      Ballantyne is right to caution that discovering a correct epistemic principle will likely be insufficient to guide thinkers. Issues of guidance will bring in a host of additional concerns, not the least of which is the extent of our doxastic control. While such a larger project is indeed interesting and worth pursuing, it is not the project of this book. Put more simply, the project of the book is to make the case that disagreement is quite epistemically significant. If correct, then we have one piece of the puzzle in place in terms of how we ought to conduct our intellectual lives, but it is only a piece, and a small one at that.

      So, my continued beliefs in deeply contentious matters is not even in direct conflict with the arguments advanced in the book. If the arguments in the book are sound, then those contentious beliefs lack a certain status, they are not synchronically epistemically justified given my awareness of the disagreements surrounding them, but none of the arguments in the book have as a conclusion that I am doing something wrong as a thinker by continuing to have those beliefs. For one thing, it doesn’t seem as though I can help but continue to believe these things. For another, the arguments concern but one type of principle that governs belief and make no conclusions about what I all-things-considered ought to believe.

      That said, I do hope that there is an effect that the views in the book can have for us. Ballantyne is surely correct that the effect is not inevitable, but endorsing the principles in the book can go a long way toward fostering intellectual humility. In an era where many are drawn to either mindless relativism or staunch dogmatism, I hope to offer a reasoned alternative. It’s true, one can be convinced by the arguments in the book and not become more intellectually humble, but if nothing else, that person has at least one more reason to be. The book also contains one argument directly pertaining to how people should act. I argue that the conclusions about the epistemology of disagreement can be coupled with a plausible principle of moral caution, and thereby give substantive moral verdicts. I gave the principle of moral caution as follows:

      If, having considered the moral permissibility of doing action A in context C, you should either believe or suspend judgment that doing A in C is a serious moral wrong and you know that refraining from doing A in C is not morally wrong, then you should not do A in C. (163)

      In the book, I applied this principle to the ethical issue of eating meat for pleasure.1 The principle of moral caution was intended as a kind of bridge between theoretical epistemological issues and applied ethics. However, even here I do not pretend that the argument has some special force to thereby make its readers into better moral beings. I hope to have correctly described a moral principle and one of its applications, but what effects this will have on the readers is just a different issue.


      Matheson, Jonathan. The Epistemology of Disagreement. Palgrave, 2015.

      ———. “Moral Caution and Disagreement.” Journal of Social Philosophy (forthcoming).

      1. This argument is now further developed in Matheson (forthcoming).



Coherence Without Conservation

In Reason and Explanation Ted Poston advances an explanatory coherentist view of justification, according to which the justification of a person’s beliefs consists in how well those beliefs fit within a virtuous explanatory system. In chapter 2, Poston argues that epistemic conservatism plays an essential role in such an account (18). Epistemic conservatism holds that, in at least some cases, holding a belief confers epistemic merit on that belief (19).1 Poston’s version of conservatism holds that “mere belief”—belief in cases of empty symmetrical evidence, where the subject lacks any evidence for or against the claim—generates justification (21–23).

There is some tension between these two ideas: On the one hand, a belief is justified by coherence within the relevant system. Poston writes, “Every belief, insofar as it is justified, requires non-experiential justification . . . from explanatory considerations” (134) and “a necessary condition for justification is that a subject’s belief is part of a virtuous explanatory system which beats competing systems” (105). On the other hand, some beliefs are justified despite—in fact because of—lack of integration with the person’s other beliefs or commitments.2

In what follows I argue that conservatism generates problems for Poston’s view, and that his project would be more cohesive and compelling if he were to jettison that component. I first describe problems with epistemic conservatism. Owing to space restrictions I focus on concerns arising specifically from Poston’s empty symmetrical evidence form of conservatism.3 I conclude by briefly suggesting Poston’s view is as attractive without conservatism. A secondary theme is that if, as seems doubtful, mere belief confers epistemic merit, this merit is not a species of justification.


The first problem is that Poston’s view entails a person could be justified in believing p, then acquire evidence supporting p and thereby become less justified—or even unjustified—in believing p. This occurs because the person with mere belief enjoys conservative justification. When she acquires evidence supporting p, she no longer faces empty symmetrical evidence, and so forfeits that justification. If the evidence is sufficiently feeble, she becomes less justified than she was before obtaining the supporting evidence. Call acquiring evidence for p and thereby becoming less justified in believing p the “implausible outcome.”

Poston himself raises this issue, and suggests two responses. Poston’s first response holds that the objection presupposes conservative justification can be quantified. He suggests conservative justification is instead “a factor that makes continued acceptance rational rather than a strength of evidence factor [that] can be represented by a number” (35). In response, first note the objection does not require that justification is quantifiable. All the objection requires is that justification (or rationality) is gradable. So long as one can be more or less justified—regardless of whether this justification can be represented by “some real number”—the problem arises. And justification is gradable. Second, suppose conservatism generates some positive epistemic status that is not gradable. Perhaps it confers licence to continuing believing, for example, where being licensed is binary.4 Even this version of conservatism generates a variant of the implausible outcome, since by gaining evidence that p one can thereby forfeit licence to believe p.

Poston advances a second response to the objection. Poston notes that “Bayesianism implies that sometimes acquiring positive evidence for p can decrease one’s overall justification for p” (36). He provides an example. Someone believes that (a) “all Fs are Gs is a natural law.” She infers that (b) “all Fs are either Gs or Hs,” where G and H are incompatible properties. Poston claims that by acquiring evidence of an F that is H the person gains information “positively relevant” to claim b, yet her justification for claim b thereby decreases. Let us suppose the example successfully illustrates that the implausible outcome can indeed arise. Precedential examples have dialectical value: they undermine the judgement that the outcome is impossible. But an example of the seemingly implausible circumstance does not suffice, since the example should illuminate the case in question. In Poston’s example, the acquired evidence for claim b undermines the original grounds for believing b. This explains why she is ultimately less justified. It is a special case.

In cases of epistemic conservatism, there is no undermining defeater. The implausible outcome arises in a general pattern: If someone with mere belief acquires any kind of supporting evidence—with any content, by any means, in any amount—the person thereby forfeits conservative justification for p. And so any evidence, if sufficiently weak, generates the implausible outcome. This generality is a problem for epistemic conservatism, since justification ought not be surrendered by gaining supporting evidence in this systematic, universal way.


David Christensen objects to epistemic conservatism. He writes,

An agent may, according to the conservative principle, correctly say “I happen to believe it—and that is part of my justification for continuing to believe it!”5

Christensen notes this assertion is problematically dogmatic. Poston draws on Gricean maxims to respond to Christensen’s objection—some assertions are infelicitous, despite being known, because they violate the conversational maxim of relevance. Poston avers that the above assertion is inappropriate because justification afforded via epistemic conservatism is always conversationally irrelevant. This is because conservative justification cannot be relied on in “challenged contexts”—contexts where whether p is “a live issue” (29). It follows that asserting “p,” where p is a mere belief, is also always infelicitous in challenged contexts (29).6 Since these assertions occur in challenged contexts, this explains why they appear problematic.

One reason these assertions are always irrelevant might be that my interlocutor can never benefit from my conservative justification; conservative justification cannot be transferred. A second reason might be that conservatively justified beliefs cannot be leaned on explicitly in any reasoning, since this renders the claim’s truth a live issue. Given that these beliefs are relatively impotent, they are almost always conversationally irrelevant.

The impotency of conservative justification is a potential weakness of Poston’s account. Since conservative justification licenses so little—we cannot assert p, deploy p in explicit reasoning, transfer the justification to an interlocutor, deploy the justification when the question of whether p arises, or use evidence for p to build on antecedent justification—why bestow the honourific “justification”? The status putatively conferred by mere belief differs so radically from normal justification, plausibly it is not a species of justification.

But perhaps a more significant problem is that Christensen’s objectionable assertion can be recast as a thought. A person might think, “I happen to believe p—and that is part of my justification for continuing to believe p.” This thought likewise seems illegitimate. But Gricean conversational maxims do not apply and so cannot explain the illegitimacy.

One potential response for Poston draws on the reasoning that underwrote the conversational irrelevance: conservative justification and conservatively justified beliefs cannot be used in challenged contexts (and perhaps in any explicit reasoning, since this renders “whether p” a live issue). Contexts where one thinks “p” or the above thought are contexts where p is a live issue, and so conservative justification cannot be deployed. Thus such thoughts seem illegitimate. Note that—parallel to Poston’s initial treatment of Christensen’s objection—this response impugns thinking “p” in challenged contexts, which might include all explicit reasoning using p, not just thinking about one’s justification for p.

One weakness of this response is that it exacerbates the concern that conservative epistemic merit, such as it is, does not warrant the label “justification.” In what sense are these beliefs justified if we cannot legitimately explicitly think them? This worry is not limited to the term “justification.” Suppose we construe epistemic conservatism in terms of “licence to believe p.” Since this licensed belief is relatively impotent—we cannot infer from p, etc.—in what sense is this a licence for a belief? (Indeed, perhaps it does not even qualify as a belief.)


The third set of worries about conservatism concern the idea of empty symmetrical evidence. There are three interwoven concerns: Whether Poston’s conception of empty symmetrical evidence is cogent, whether there are instances of empty symmetrical evidence, and if so whether such beliefs have any epistemic merit. In what follows my comments should be understood as invitations for clarification, rather than as objections.

Poston’s central examples of empty symmetrical evidence are hinge propositions (21–22). In Wittgensteinian epistemology, hinge propositions are the background against which we inquire. In On Certainty examples of hinge propositions include that the world has existed for a long time, 12 x 12 = 144, and that one has hands. Poston adds claims about the reliability of memory, stability of meaning, and object permanence (22), which are “fundamental background beliefs” (127).7

But one might doubt that these are commitments without evidence. According to most conceptions of evidence we possess evidence for these kinds of claims. If there is no evidence, epistemologists must explain why people believe these claims rather than their negations and what underwrites the broad agreement. If evidence can be gained via coherent explanations, as coherentists maintain, then plausibly we have evidence supporting, for example, the reliability of memory. It is not clear these examples are best understood as cases of empty symmetrical evidence.

My second query concerns whether there is tension between Poston’s claims that for some beliefs we have justification without evidence and his treatment of various foundationalist claims. In chapter 5 Poston challenges the doctrine of “the given” by arguing no empirical belief is properly basic because even beliefs about the putative given rely on integration with background beliefs.8 In chapter 6 Poston criticises the possibility of self-evident a priori claims, such as everything is self-identical and no surface can be simultaneously uniformly red and blue (133). Poston’s approach denies the possibility of autonomous rational insight, whereby justification for a belief is “epistemically autonomous . . . dependent on nothing beyond itself for its justification” (135). Poston argues “every belief, insofar as it is justified, requires non-experiential justification. But the justification comes not from some mysterious faculty that supposedly directly grasps necessary features of reality, but rather from explanatory considerations” (134) and emphasises that every justified belief is embedded in a coherent explanatory story (135–39). He writes, “[The putatively self-evident a priori belief] does not stand alone and does not justify apart from its impressive coherence with other elements of the story” (136). Evidence, justification, and reasons are, apparently, ubiquitous. But if evidence is everywhere, perhaps similarly there is evidence for so-called hinge propositions.

Poston’s positive view is thoroughly coherentist. Throughout the book, Poston emphasises that “a belief is justified only if it is part of a virtuous explanatory system which beats its competitors. A belief can be a part of such a system by either explaining, being explained, or being a part of an explainer or explainee” (102).

Given Poston’s arguments that putatively self-evident a priori claims and the putative “given” are justified only in virtue of their integration within a coherent web of beliefs, and given his emphasis on how the justification of a belief is a matter of explanatory coherence, it seems discordant that Poston posits commitments that are believed and justified without evidence. Whilst not necessarily in contradiction, since justification could be a disjunctive kind, for example, Poston’s commitment to coherentism about justification seems a strange bedfellow for epistemic conservatism, which effectively endorses that some beliefs are “epistemically autonomous . . . dependent on nothing beyond itself for its justification.” One might resolve this apparent conflict by positing that conservative justification is substantially different from normal justification. But this re-raises the question of whether conservative justification is a species of justification at all.

Bluntly, if the belief that the world is very old can be justified autonomously without evidence, why can’t the belief that everything is self-identical? I suggest Poston’s view would be more cohesive if conservatism is jettisoned. In the final section I briefly suggest how he might retain the advantages of his view without conservatism, but first I articulate one further concern about his conception of empty symmetrical evidence.


In what follows I question whether empty symmetrical evidence is possible. Poston notes that prior to consulting the result, a belief that Johnson won the lottery does not qualify as empty symmetrical evidence because one possesses evidence that it is unlikely Johnson won (21). And prior to looking, a belief about the outcome of a coin toss does not qualify as empty symmetrical evidence because “one will shortly have positive evidence” (21). In personal communication, I asked Poston: if I pocketed the coin without checking the result, would this qualify as empty symmetrical evidence? In this case, evidence is not forthcoming. Poston replied that it would not, since I have “relevant evidence”; “evidential principles are still in play.”9 But here a tension seems to arise.

To see the worry, suppose Carrie is ignorant of Chinese geography. Consider the claim that there are 18 provinces in China. This case is not symmetrical: It is very likely to be false. It resembles the lottery belief described above. Suppose instead Carrie believes there are between 14 and 26 provinces in China. This claim, with its broadened numerical range, is more likely to be correct. Does it qualify as empty symmetrical evidence? Unlike the lottery example, and the first claim about China, the belief is not outlandish. It might well be correct; perhaps it is symmetrical. But the very feature that underwrites its symmetry—the parity in the chances—leads it to resemble the pocketed coin toss case. Carrie has “relevant evidence.” Evidential principles are still in play: It is not empty. If the pocketed coin toss case does not qualify as empty symmetrical evidence, then plausibly neither should this belief.

Poston provides specific reasons why each mundane case does not qualify as empty symmetrical evidence.10 Perhaps the reason none of these everyday cases qualify is not the specific features of each case, but a general tension in the idea of empty symmetrical evidence: The belief must be—in order to be empty—one that we have no relevant evidence about, yet must nonetheless be sufficiently plausible to be symmetrical. There may be a tension between emptiness and symmetry.11

In chapter 4 Poston describes a putative case of empty symmetrical evidence. Sally believes the Battle of Hastings occurred in 1066. She learnt this from a reliable source, but has since forgotten the original acquisition (97). Poston says,

To make this case one of empty symmetrical evidence we need to engage in some fiction. Suppose all historical records relevant to the date of the Battle of Hastings are destroyed and Sally is unable to consult anyone else. In this fictional case, all she has to go on is her conviction that she has it right. (98n48, emphasis mine)

Sally is supposed to have no evidence. (I think the legitimate, albeit forgotten, genealogy of her belief is evidentially relevant.) But if she has no evidence, as Poston claims, presumably we can amend the case so she believes the battle occurred in 1031 and was never told about any battle date. She simply has the conviction that she is right. But this case resembles the lottery and “18 provinces” examples. It is not symmetrical; the belief is likely to be false.12 If conservatism renders Sally’s belief justified, conservatism is implausible. But the deeper problem is that Poston’s particular version of conservatism appears to rest on a shaky foundation: empty symmetrical evidence seems unviable. If a belief is empty, it is not symmetrical; if symmetrical, it is not empty.13


In conclusion I highlight some of Poston’s motivations for conservatism, and briefly suggest how his account nonetheless thrives without conservatism.

One of Poston’s motivations for conservatism is to capture the perspectival nature of epistemic justification, namely that the justification of a subject’s beliefs depends on her perspective (39). But coherentism without conservation preserves the perspectival nature of justification. It is the person’s beliefs, coherence-making relations, history, and understanding that determine the justificatory status of her beliefs. Given this, it is not clear how much “additional” perspectival character conservatism provides: Conservative justification may apply only to a small set of beliefs, and is arguably redundant once reasoning begins. In cases of empty symmetrical evidence, because such cases are prior to reasoning, one’s personal commitments and epistemic agency arguably do not play a substantial role. One’s perspective is more significant once reasoning begins. So plausibly the perspectival character of justification is best captured by those beliefs where conservatism does not apply.

Poston posits that conservatively justified beliefs are later vindicated by their role in a coherent explanation.14 If the question of whether p arises, conservative justification is impotent (41). But later explanatory connections generate “better justification” for p (38). My suggestion is that conservative justification is dispensable; Poston could instead hold that mere beliefs are not justified until vindicated by coherence. The beliefs must be later vindicated by coherence anyway, so why not posit that all justification arises from coherence? The resulting view is more unified, and avoids problems generated by conservatism. Poston’s view is an account of epistemic justification or reasonableness. It is compatible with his account that there are other dimensions of epistemic normativity, such as understanding-why, knowledge, exemplification, and explanation that require an externalist anchor to reality in addition to coherence amongst beliefs.15

Poston concludes his defence of epistemic conservatism,

Conservatism is an important epistemological doctrine. Anyone attracted to a non-skeptical internalist epistemology must come to grips with the ineliminable role of background beliefs for epistemic justification. Following Neurath’s famous metaphor, the entirety of our system of beliefs is like a raft that we rebuild at sea. We seek to continuously improve the raft by adding and subtracting new material but we must always stand on some existing timber.

But we should keep asunder two claims: The first, coherentism, holds a belief is justified in virtue of its position in a coherent web of beliefs. The second, conservatism, holds a belief is justified in virtue of merely being believed. The former is a plausible claim. The latter should be jettisoned.16


Christensen (1994). “Conservatism in Epistemology.” Noûs 28(1):69–89.

Foley (1983). “Epistemic Conservatism.” Philosophical Studies 43:165–82.

Fumerton (2008). “Epistemic Conservatism: Theft or Honest Toil?” Oxford Studies in Epistemology, Gendler and Hawthorne (eds.), OUP, 63–86.

Poston (2014). Reason and Explanation. Palgrave Macmillan.

Sellars (1963). Science, Perception and Reality. Routledge & Kegan Paul.

Vahid (2004). “Varieties of Epistemic Conservatism.” Synthese 141:97–122.

Wittgenstein (1969). On Certainty. Harper and Row.


  1. For helpful taxonomies of epistemic conservatism, see Valah (2004) and Fumerton (2008). Fumerton (2008) and Christensen (1999) caution against conflating conservatism with a “spurious” version, where one uses a track record, or some other evidence that one’s beliefs are reliable, to confer merit on the belief.

  2. As I explain below, once a belief becomes integrated, it forfeits conservative justification.

  3. For general objections to epistemic conservatism, see Foley (1982), Christensen (1999), Fumerton (2008).

  4. This proposal might be what Poston intends by “a factor that makes continued acceptance rational”; being rational might be ambiguous between a gradable and binary property.

  5. Christensen (1994:69) quoted in Poston (2014:28).

  6. Poston writes: “In any context in which the question of whether p is true is a live issue, our adherence to the cooperative principle makes one’s belief that p entirely irrelevant to the discussion of whether p is true. In such contexts the assertion that “I believe p” creates the false expectation that one’s believing p is relevant for the purposes at hand.” (29, emphasis mine) But note a similar issue arises for the mere assertion “p.”

  7. Note that whether a claim is a hinge proposition may change with context. Owing to lack of space, I leave aside discussion of whether according to Wittgensteinian epistemology hinge propositions are beliefs, endorsed without evidence, justification-apt, or ineligible for epistemic evaluation.

  8. Sellars (1963).

  9. Ted Poston, personal communications, March 2016. I greatly appreciate Poston’s helpful conversations about his ideas.

  10. By “mundane” I mean a case that does not rely on Wittgensteinian epistemology or some specific or controversial conception of evidence.

  11. One potential response is to weaken the requirements for emptiness, such that the pocketed coin toss case qualifies as empty symmetrical evidence; Poston could hold that balanced statistical evidence or having “evidential principles still in play” is consistent with empty symmetrical evidence. By rendering the coin-toss belief conservatively justified, conservatism generates an implausible result. But this route might resolve the potential tension in the conception of empty symmetrical evidence.

  12. Perhaps Sally’s belief is rendered symmetrical by her strong conviction, and thereby differs from the lottery and “18 provinces” claims. But this suggestion underscores the implausibility of conservatism: Her dogmatic conviction alone transforms a deeply improbable claim into an instance of empty symmetrical evidence, and thereby renders it conservatively justified. If epistemic conservatism vindicates Sally’s belief, it vindicates the belief that Johnson won the lottery, despite no evidence, so long as the believer is sufficiently confident in his belief. This version of conservatism vindicates dogmatism, but only when the belief is strong despite lack of evidence.

  13. Another worry about the idea of empty symmetrical evidence concerns whether it renders conservatism ad hoc. It seems somewhat ad hoc that belief is justification-conferring in cases of empty symmetrical evidence, but not in cases of symmetrical but non-empty evidence, when evidence is forthcoming, when we can apply relevant evidential principles, or indeed, in general for all belief.

  14. See, e.g., pp. 38, 41, 67, 127.

  15. I do not have space to develop this point here, but the observation about other epistemic states can help mitigate the alternative systems objection. It is less worrisome that other systems are equally justified, for example, given that justification is only one dimension of epistemic normativity.

  16. I am grateful to Bob Beddor, David, Black, Will Fleisher, Jon Garthoff, Ted Poston, Duncan Pritchard, Daniel Rubio, and Ernie Sosa for helpful comments and discussions about these issues.

  • Avatar

    Ted Poston


    Coherence and Conservation: A Response to Gardiner

    I am grateful for Gardiner’s excellent and challenging comments on my book.1 I cannot hope to adequately address all of the objections she raises. Instead I will discuss the reasons I think epistemic conservatism is required for a plausible coherentist view and then I will discuss the core idea behind conservatism. My hope is that within the context of a properly formulated and motivated conservatism some of the most pressing concerns Gardiner raises will appear less troubling than initial appearances.

    I. Coherentism Without Conservatism

    One of Gardiner’s most interesting insights is that an explanationist view similar to mine but trimmed of epistemic conservatism is more plausible. Let us refer to this view as trimmed explanationism. Gardiner claims that trimmed explanationism is more plausible than my conservative version of explanatory coherentism. Her thought is that epistemic conservatism has a number of implausible consequences and it’d be better overall to drop a feature of a theory when the trimmed theory does just as well. I address the particular concerns about epistemic conservatism in the second half of this response. Let us then consider the claim that trimmed explanationism does just as well as conservative explanationism.

    To assess this claim, we need first to consider the nature of trimmed explanationism. The view is this. The justification of a subject’s belief that p depends entirely on the explanatory virtues of that belief in relation to the subject’s other beliefs. The virtues do not include epistemic conservatism; rather they are exhausted by simplicity and power.

    Trimmed explanationism, like my conservatism explanationism, is a specific instance of a coherentist epistemology. In general, a coherentist view holds that the justification of a subject’s belief that p depends entirely on the coherence relations of that belief in relation to some other system S. A coherentist must specify the nature of coherence and the relata of the coherence relation. In my book I argue that explanatory coherence, particularly the virtues of simplicity, power, and fit with background beliefs, are plausible candidates for coherence relations. Moreover, I argue that a plausible choice of S is a subject’s beliefs and experiences (where we understand beliefs and experiences as two ways of mentally hosting assertive representational content).

    Trimmed explanationism mirrors my view minus conservatism. It holds that the nature of coherence is given by the specific explanatory virtues of simplicity and power. Moreover, it mirrors my view in holding that the second relatum of the coherence relation is the set of the subject’s beliefs and experiences. It is reasonable to expect that in many cases trimmed explanationism and conservative explanationism will agree about cases. If they were in complete agreement then trimmed explanationism would be the better view.

    There are two reasons to favor conservative explanationism over trimmed explanationism. First, trimmed explanationism lacks a natural answer for why the second relatum of the coherence relation should be restricted to the subject’s beliefs and experiences? If the explanatory virtues are only simplicity and power then a trimmed explanationist view seems unable to resist the thought that justification depends on the simplest and most powerful system (full stop). It does not matter who’s system it is. If the only virtues that matter for justification are simplicity and power then the justification of a subject’s belief that p depends on the fact about which system is the simplest and most powerful. It need not matter whether this system conflicts widely with the beliefs of a subject. Copernicus’s view of the universe was simpler and more powerful than Ptolemy’s. But it conflicted with the then accepted Aristotelian theory of motion. The simplicity and power of Copernicus’s view were not enough to make it reasonable to accept until Galileo successfully overturned the old theory of motion.

    The argument in the previous paragraph is too quick, but the way in which it is too quick doesn’t help trimmed explanationism. The virtues of simplicity and power are contextual. Presumably, the simplest and most powerful view would be a naive Parmenidean view according to which, appearances be damned, reality is one. On the naive view, we don’t attempt to explain away appearances of diversity, we reject them. The resulting view is very simple and perfectly powerful (there is nothing that is not completely explained). It is right to complain that the naive Parmenidean view is not really powerful because it achieves it success by rejecting the data to be explained. But this is precisely the point I want to raise: simplicity and power are contextual virtues that require data. A view should be simple relative to the data to be explained. Where does the data come from? I’ve argued in chapters 5 and 6 the traditional appeals to the empirically given or the rational a priori do not succeed. The data comes from things that we accept, from belief. So trimmed explanationism requires conservatism. It requires that there be some data such that large scale theories are assessed for simplicity and powerful in light of the data. If the argument in this paragraph is correct then, contrary to Gardiner’s claim that conservatism and explanationism are in tension, there is a deep concord between the two.

    As I see it trimmed explanationism is either less plausible than conservative explanationism or equivalent to it. First, assuming that one can really do without conservatism, trimmed explanationism doesn’t constrain justification in the right kinds of ways because the justification of a belief would not track facts about the data that one accepts, where the inclusion of that data makes the resulting view less simple and/or less powerful. Second, trimmed explanationism really requires conservatism because the virtues of simplicity and powerful are contextual, requiring a context of accepted data. I should add that if the data has no presumption in its favor then I don’t understand why such data should constrain rational belief.

    II. Core Conservatism

    Gardiner raises a number of formidable objections to conservatism and my specific formulation of it. Some of the challenges are raised by other commentators and, because of space limitations, I will address some of those concerns elsewhere. As I say in my book, epistemic conservatism is not a popular doctrine. A conservative epistemologist is a subversive radical epistemologist. I continue to think that epistemic conservatism is poorly understood in part because of the significant changes it requires to the way we conceive of some fundamental epistemic notions. I attempted to address this in my second chapter, but clearly more work needs to be done.

    Let me step back and get at the core conservative claim. The basic idea is that a subject’s belief that p makes it, prima facie, rational to continue to believe p. As I specify in the book, the core conservative idea does not license dogmatism. Epistemic conservatism is naturally aligned with a respect for sound principles of evidence. How so? As I see it, belief is teleologically ordered to knowledge. That is, the goal of belief is knowledge. A belief that is not knowledge is normatively defective. It follows from this that a subject who believes p without knowing p is under a general norm to inquire further. I understand conservatism as fitting naturally with this norm of inquiry that is closed when knowledge is achieved.

    A consequence of this conception of belief and inquiry is that any justification that comes from belief itself does not permit closure of inquiry. Thus, to the extent that one has some level of justification from epistemic conservatism it is normatively lacking. In the book, I attempted to get at this normative defect in conservative justification by specifying that a belief that is only justified by conservative considerations does not have the normative strength to be used (all by itself) in assertion and reasoning. I conceive of conservative justification in a dynamic rather than a static setting. A belief held in the state in which a subject has no evidence for it has two properties: first, the subject has a right to maintain that belief for a time, but the subject has an intellectual obligation to inquire further. If inquiry does not support the belief—in terms of my view, if the belief cannot achieve explanatory coherence—then the belief ought to be given up. For the resulting system of beliefs is simpler without the one that cannot be incorporated.

    Corresponding to the dynamic setting and weakness of conservative justification, is the idea that the justification that a belief possesses by conservatism is enhanced by coherence relations. Thus the kind of justification that matters for knowledge requires this enhanced explanatory coherence justification. Gardiner is right to highlight the apparent tension between my conservative position—a belief has some justification in absence of coherence relations—and my coherentist position—a belief only has justification by its coherence relations. The tension is resolved by acknowledging that the justification that comes from coherence relations is the kind of justification that matters for knowledge.

    I hope that these remarks remove some of the implausibility epistemologists find in epistemic conservatism. Let me close with a brief remark about agency and belief. There is human freedom in action. A subject is permitted to choose her actions if they do not conflict with moral requirements. Similarly, with belief, a subject is free to have any belief that does not conflict with her epistemic obligations (note: this doesn’t require doxastic voluntarism). As I see it the key question for a non-skeptical epistemology is whether we are free in belief to have some basic starting points which then aim for explanatory coherence. Explanatory coherence is demanding so that not every starting point is vindicated. But if some starting points can be vindicated then that is indeed an epistemic success. If human freedom does not extend to belief then we lose the perspective that knowledge is a well-earned human achievement that comes from humble beginnings.

    1. Thanks to Kevin McCain for helpful comments on this essay.



Half-Full Glass: Poston’s Epistemic Conservatism

Poston argues that a belief that P is propositionally justified if, and only if, it is part of a system of beliefs that has a higher explanatory value than the alternatives. This account is intended to be a coherence theory that is both evidentialist and anti-foundationalist. As an evidentialist theory, it is supposed to require that the justificatory status of a belief is determined by the evidence available to a person. Unlike foundationalist theories, it is supposed to require that persons have reasons for their justified beliefs. In fact, however, Poston’s theory is a version of foundationalism since it does not require that justified beliefs be supported by reasons. His theory is also incompatible with evidentialism. The problem is with his epistemic conservatism, the view that in the absence of any relevant evidence propositions are justified if they are believed. Poston’s conservatism makes it hard to distinguish his theory from a powerful version of skepticism.

A traditional foundationalist theory holds that every justified belief must either be basic—justified independently of reasons for belief that are provided by other beliefs—or be the first component of a finite sequence of reasons for belief that terminates with a basic belief. Traditional coherence theories hold that there are no basic beliefs but that every justified belief must be the first component of a finite sequence of reasons for belief. So justification requires circles of reasons. Infinitist theories hold that every justified belief must be the first component of an infinite sequence of reasons for belief. Poston claims that his holistic coherence theory is an alternative to these traditional anti-skeptical responses to the regress problem:

Justification is not foundational because any specific claim requires a reason. Justification does not require an infinite regress because holistic support does not require an infinite number of non-repeating claims. And justification is not circular because holistic support is different from circular arguments. (2)1

Poston develops his theory in response to a version of the regress problem due to Bergmann. This problem is an argument for skepticism that proceeds from the negation of a critical foundationalist principle, F, according to which some beliefs can be non-inferentially justified:

  1. A belief can be justified only if it is inferentially justified. [i.e. ØF]
  2. A belief can be inferentially justified only if the belief from which it is inferred is a justified belief.
  3. Therefore, a belief is justified only if it is justified via logically circular reasoning or it is justified via an infinite chain of reasoning. [from 1 and 2]
  4. No beliefs can be justified via logically circular reasoning.
  5. No beliefs can be justified via infinite chains of reasoning.
  6. Therefore, none of our beliefs are justified. [from 3, 4, and 5] (Bergmann: 185, quoted by Poston: 65)[/NL]

Poston claims that the mistake in this argument is that 3 does not follow from 1 and 2 because “a belief can be justified by its relations to other beliefs without approving either logically circular reasoning or an infinite chain of reasoning” (65).

To evaluate Poston’s response, we need to understand premises 1 and 2, and that requires an account of inferential justification. Bergmann says that a belief is noninferentially justified if, and only if, “it is justified but not in virtue of being inferred from or based on another belief” (Bergmann: 184, quoted by Poston: 64). Given this, a belief is inferentially justified if, and only if, it is justified in virtue of being inferred from or based on another belief. It is natural to take this to be the view that inferentially justified beliefs are justified because persons have reasons for them:

For all P1, S’s belief that P1 is inferentially justified if, and only if, S’s belief that P1 is justified because there is a P2 such that S believes P2 and S’s belief that P2 provides S with a reason for believing P1.

If so, then 1—all justification is inferential—is this principle:

1¢         For all P1, S’s belief that P1 is justified only if there is a P2 such that S believes P2 and S’s belief that P2 provides S with a reason for believing P1.

Likewise, 2 is naturally taken to be this principle to the effect that reason-providing beliefs must be justified:

2¢         For all P1 and P2, S’s belief that P2 provides S with a reason for believing P1 only if S’s belief that P2 is justified.

Taken together 1¢ and 2¢ entail:

3¢         For all P1, S’s belief that P1 is justified only if there is a P2 such that S believes P2, S’s belief that P2 is justified, and S’s belief that P2 provides S with a reason for believing P1.

3¢ is true, however, if, and only if, every justified belief requires either a circle or an infinite regress of reasons for belief. So if this interpretation of the regress argument is correct, 1 and 2 do entail 3. Poston must, therefore, reject this interpretation.

He does. Poston takes Bergmann’s account of noninferential justification to be akin to Alston’s account of immediacy according to which a proposition is immediately epistemized (given positive epistemic value) for a person S just in case it is “epistemized by something other than some relation this belief has to some other epistemized belief(s) by S” (Alston: 75, quoted on Poston: 65, my emphasis). Given this account, a belief is inferentially justified just in case it is justified because it stands in some epistemically relevant relation to some other justified belief. Even if every justified belief that P must stand in some epistemically relevant relationship to another justified belief, it does not follow that this relationship is a relationship between the belief that P and a reason for that belief. For on this interpretation of inferential justification, premise 1 says only this:

1¢¢        For all P1, S’s belief that P1 is justified only if there is a P2 such that S believes P2 and S’s belief that P2 stands in some epistemically significant relationship to S’s belief that P1.

Likewise, premise 2 would say:

2¢¢        For all P1 and P2, S’s belief that P2 stands in some epistemically significant relationship to S’s belief that P1 only if S’s belief that P2 is justified.

It is clear that 1¢¢ and either 2¢ or 2¢¢ do not jointly imply that justification requires circles or infinite regresses of reasons for belief. So it might seem that if a coherence theory implies that 1¢ is false but that 1¢¢ is true, it might be an anti-skeptical solution to the regress problem that is an alternative to traditional foundationalism, coherentism, and infinitism. This is Poston’s strategy. He argues that a belief is justified for a person by being included with other justified beliefs in a sufficiently virtuous explanatory system. The problem is that Poston’s view avoids requiring circles or infinite regresses of reasons by being incompatible with the promised view that justification requires a reason for belief. As a result, his theory is a version of foundationalism not an alternative to it. Let me show how this is so.

Poston’s theory holds that a belief that P is justified if, and only if, it is part of a system of beliefs that is at least tied in explanatory value with alternative systems that agree about P and has more explanatory value than alternative systems that do not agree about P:

Ex J¢¢: S has justification for believing p if and only if p is a member of a sufficiently virtuous explanatory system, E, and E is more virtuous than any p-relevant competing system E¢. (90)

The most important explanatory values are “conservativeness, explanatory power, and simplicity” (14) and an explanation is conservative to the extent that it conflicts with as few of a person’s prior beliefs as possible. Poston holds that a preference for explanations that do not conflict with a person’s prior beliefs has more than merely pragmatic or heuristic value because epistemic conservatism is true:

[I]n the special evidential situation in which a subject believes p but lacks any relevant evidence for p then a subject has some justification for maintaining her belief that p. (21)

Poston claims that:

Conservatism lays at the heart of any feasible explanatory coherentist view. To the extent that anyone is justified in believing anything it is because of a host of background beliefs. Background beliefs together with the other explanatory virtues provide the ultimate material for justification. I do not see any way to resist a thoroughgoing skepticism without these background beliefs having some level of justification simply in virtue of being believed. (18)

Poston thinks that his epistemic conservatism is compatible with the view that justification is a matter of being part of the best available explanatory system because conservatism is an explanatory virtue and this is so because, all else being equal, one explanation is better than another to the extent that it requires fewer changes to a person’s prior beliefs.

There are three ways a belief can be a part of an explanatory system: it can be a belief about an explained but non-explaining state of affairs, it can be a belief about an explaining state of affairs, or it can be a belief about a state of affairs that is neither explained nor explaining. Not all beliefs in an explanatory system need to be either beliefs about explaining or explained states of affairs. A system of beliefs has explanatory value just in case the explaining propositions the system contains are sufficiently powerful, simple, and conservative compared to the alternatives. Poston’s epistemic conservatism plays at least two important roles in his theory. First, it provides him with an account of how we can be justified in believing some explained but non-explaining propositions. Second, it provides him with an account of how we can be justified in believing explaining or non-explaining propositions for which we have no evidence, including important “hinge propositions” (21−22) for which we cannot have evidence. Hinge propositions are important because they are needed for the conduct of inquiry but cannot be supported by evidence. They include such things as that meanings are stable and that induction is reliable. The distinguishing feature of hinge propositions is that the only arguments we could have for them are rule circular (22).

One way a belief for which we have no reason contributes to the explanatory virtue of the system as a whole is by providing a target for explanation. Only facts require explanations so we must have a stock of justified beliefs about purported facts in order to begin the process of constructing an explanatory system. Another way a belief for which we have no reason contributes to the explanatory virtue of a system of beliefs is by not conflicting with the explanations the system does provide. For the more beliefs that are left in place by an explanatory belief, the more conservative the explanatory system.

We are now in a position to see how Poston’s theory is a version of foundationalism. For his account implies that some beliefs are basic in the sense that they can be justified for us despite the fact that we have no reason for believing them and, given Poston’s assumption that circles and infinite regresses of reasons are not required for justification, Ex J¢¢ implies that every justified belief is either basic or is the first component of a finite sequence of reasons for belief that terminates with a basic belief. Poston’s theory is not a standard version of foundationalism because the basic beliefs required by his system can be about anything and can be produced by an indefinite range of belief-forming processes (62). The holistic explanatory properties of a system of beliefs explain why the individual beliefs in the system are justified but—this is the critical point—the system must contain some beliefs that are justified even though the believer has no reasons for them. That the explanatory goodness of a system provides a sufficient condition for a belief’s being justified does not imply that a person has a reason for that belief. Poston could avoid foundationalism by requiring that a belief is justified for a person only if it is supported by a belief about the explanatory system. This requirement would, however, be incompatible with his conservatism, be implausibly strong, and be dangerously close to implying the sort of skepticism his theory is designed to avoid.

Contrary to what he argues, Poston’s epistemic conservatism also makes his theory incompatible with evidentialism. Evidentialism is the view that the justificatory status of a doxastic attitude is determined by the evidence that persons have. Evidentialism is expressed by this principle about the supervenience of justification on evidence:

ES: The epistemic justification of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that person has at the time. (92)

The principle ES is equivalent to this principle:

Necessarily, persons alike with respect to the evidence available to them must be alike with respect to the justification of their doxastic attitudes.

This principle entails the following difference principle about justified beliefs:

Diff:     For all P, if a person S1 is justified in believing P and another person S2 is not justified in believing P, then there is a relevant difference between the evidence available to S1 and the evidence available to S2.

Poston’s conservatism entails that Diff is false, hence that ES is false. So Poston’s conservatism is incompatible with evidentialism.

Let me explain. Poston’s epistemic conservatism implies that there can be two persons S1 and S2 who are alike with respect to the evidence available to them but that S1 is but S2 is not justified in believing a proposition P, contrary to Diff and, therefore, contrary to ES. For suppose that two persons S1 and S2 are alike with respect to all of the evidence available to them, that this evidence is not relevant to a given proposition P, and that S1 believes P but S2 does not believe P. Poston’s version of conservatism implies that S1 is justified in believing that P but S2 is not justified in believing P even though S1 and S2 are alike with respect to the evidence available to them. This is incompatible with Diff. So Poston’s conservatism makes his theory incompatible with evidentialism. Poston might think that his system is compatible with evidentialism because he takes ES to be equivalent to a principle to the effect that having supporting evidence is sufficient for justification:

For all P, if a person S has total evidence E and E supports P, then S is justified in believing P.

This principle implies that any two persons alike with respect to evidence that supports P are both justified in believing that P but leaves it open that there are sufficient conditions for justification that do not require supporting evidence. The evidentialism expressed by ES, however, implies that justificatory differences require differences in evidence. Poston’s epistemic conservatism is not compatible with this.

Well, so what? Why does it matter that Poston’s theory is a nonstandard version of foundationalism or that it is incompatible with evidentialism? These things matters for several reasons. Poston’s view implies that it is epistemically OK—epistemically good, permissible, virtuous, or in some other way appropriate—to believe propositions that are epistemically arbitrary from a person’s own point of view. For it is a consequence of Poston’s theory that it is OK to believe propositions for which one has no evidence and for which one has no reasons provided that one has no evidence or reasons relevant to those propositions at all. An attractive feature of this sort of view is that it allows persons to be justified in believing lots of useful propositions for which they have no evidence such as the proposition that induction is reliable. A problem is that Poston’s theory cannot rule out being justified in believing lots of extravagant propositions for which persons also lack evidence or reasons: that the number of stars is even, that there are 25 undetectable ghosts in the room now, or even that induction is not reliable, for example. Indeed, since beliefs for which we lack evidence or reasons contribute to the explanatory goodness of our systems of beliefs by not being ruled out by the explanations we have, it seems to be a consequence of Poston’s theory that, all else being equal, it is a good thing to maximize the number of beliefs for which we have no evidence or reasons since doing so would increase the conservativeness of our explanatory systems.

One of the attractive things about evidentialism and about the related view that justification requires reasons for belief is that these imply that beliefs that are epistemically arbitrary from our own points of view are not justified. As a result of his conservatism, Poston’s view implies that justified beliefs can be arbitrary from our own points of view. Because of this, it is hard to distinguish Poston’s view from a version of skepticism according to which it might be practically valuable to have a system of beliefs with the explanatory virtues that Poston recognizes, but that we should not suppose that this makes our beliefs epistemically valuable. This kind of skepticism is motivated by the regress problem and holds that our beliefs lack epistemic value because we cannot have the kind of evidence or the kinds of reasons for them that epistemic value requires. Unlike some versions of skepticism but like Poston’s explanatory coherentism, it does not hold that we can or ought to suspend judgment about propositions for which we lack good evidence or good reasons. Since one of Poston’s arguments for conservatism is that it is the only alternative to skepticism (39), I wonder what the difference is between this kind of skepticism and Poston’s view. Perhaps the difference is that although this sort of skepticism takes our epistemic glass to be half empty, Poston takes it to be half full.



Alston, William. 1983. “What’s Wrong with Immediate Knowledge?” Synthese 55: 73−95.

Bergmann, Michael. 2006. Justification Without Awareness. New York: Oxford University Press.

Poston, Ted. 2014. Reason and Explanation: A Defense of Explanatory Coherentism. New York: Palgrave.

  1. All citations are to Poston unless otherwise indicated.

  • Avatar

    Ted Poston


    Belief, Evidence, and Knowledge: A Response to Cling


    I thank Andy Cling for these careful and insightful comments. Cling effectively summarizes many of the motivations and arguments I give for my explanationist view. He argues that on several important dimensions my view does not live up to its promises. In particular, he charges that the version of explanatory coherentism I defend is not a form of evidentialism and, moreover, it is a kind of foundationalist skepticism. In the following I aim to answer these important claims.

    Let us begin with Cling’s argument that explanationism is not a form of evidentialism. Apart from terminological debates, this issue is important because it is a natural judgment that a person’s evidential position completely determines the justificatory status of their beliefs. If two subjects who are identical evidentially may differ in what attitudes are justified for them then a person’s evidential position does not completely determine the justificatory status of their beliefs. As I discuss in my book evidentialism is captured by the following thesis:

    (ES) The epistemic justification of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that person has at that time. (92)

    Cling observes that (ES) implies a difference principle.

    (Diff) For all P, if a person S1 is justified in believing P and another person S2 is not justified in believing P, then there is a relevant difference between the evidence available to S1 and the evidence available to S2.

    The problem is that epistemic conservatism implies that (Diff) is false because two persons may be in identical evidential situations and yet if S1 believes p and S2 does not then S1 is justified in believing P but S2 is not. Epistemic conservatism then does not fit with an evidentialist epistemology.

    In my book I discuss the problem that conservatism poses for evidentialism (see 94–95). I am grateful to Cling for pointing out the need to revise that discussion. The issue is how to understand the relevance of belief in the special state of empty evidence. One simple solution is to take the belief that p to be very weak evidence for p. I do not favor this solution because evidence is the kind of thing that can be brought to bear on a conversation over whether p is true. Even if experiential evidence cannot always be shared, one can tell another person about the experience. In this case one says that one has some sort of experience, but one cannot show the experience. Belief does not have this character. For reasons I give in the book, a subject’s belief that p is not the kind of thing that bears on such a conversation. Rather a subject’s belief that p is part of her perspective which in the special state of empty evidence she has a right to maintain. As I explain in my response to Gardiner, belief is teleologically ordered to knowledge and so if a belief cannot achieve coherence then the belief must be surrendered. Even so, on my view a subject does have some justification for maintaining her belief in the state of empty evidence and an otherwise identical subject lacking such a belief would not have that justification for that proposition.

    Even if (ES) is inconsistent with conservatism, there is a version of (ES) that is not. (ES) is a thesis about epistemic justification and if there are kinds of justification then there are different versions of (ES). A coherentist may hold that knowledge-level justification requires coherence and so the knowledge-level justificatory status of a subject’s beliefs is entirely determined by the evidence. Weak justification, the kind that comes from conservatism, is not entirely determined by the evidence. So, explanationism is consistent with the following principle:

    (KES) The knowledge-level justification of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that person has at that time.

    Explanationists and evidentialists can both affirm (KES). Moreover, the debates over evidentialism are unaffected by substituting (KES) for (ES). All the standard non-evidentialist views will equally deny (ES) and (KES).

    There is a deeper agreement between evidentialists and explanationists that deserves stressing. Both views affirm mentalism. Mentalism is the following thesis:

    (S: The justificatory status of a person’s doxastic attitudes strongly supervenes on the person’s occurrent and dispositional mental states, events and conditions. (91)

    Belief is a mental state and so epistemic conservatism doesn’t depart from evidentialism over S. The mentalist thesis (S) is a natural way to stress the idea that epistemic justification supervenes on a subject’s perspective. Any two subjects who are mentally identical are identical with respect to which attitudes are justified for them. In consequence, epistemic conservatism is not inhospitable to evidentialism with respect to mentalism.

    Cling contends that conservatism poses problems elsewhere for my view. He argues that conservatism is incompatible with anti-foundationalism and tends to imply a kind of skepticism. I first address the worry over foundationalism. Cling helpfully lays out my strategy for arguing for a unique response to the regress argument. In chapter 3 I argue that foundationalism is best viewed as a combination of views. First, there is a view about justification in absence of any reasons. Second, there is a view about inference from basic beliefs, what I call “direct ampliative inference.” I argue that a key coherentist argument—the argument against first philosophy—targets the foundationalist view about direct ampliative inference, not the foundationalist view about justification apart from reasons. In response to this argument, I formulate a view of reasons on which reasons require a framework of justified commitments. Cling is right that my view allows for some weak justification apart from reasons. But, because knowledge requires reasons, my view does not allow that one can have knowledge-level justification apart from reasons.

    It is not evident to me that there is any particular problem with this combination of views. Cling says that my view implies “every justified belief is basic or the first component of a finite sequence of reasons that terminates with a basic belief.” Taken out of the context of my views in chapters 2 and 3, this statement is misleading. To the extent that a justified belief is basic, its justification is insufficient for knowledge. Given my INUS account of reasons, a belief that has the kind of justification necessary for knowledge is supported by a large framework of justified beliefs. Reasons can be given but those reasons depend on a body of other justified beliefs. So, on my explanatory coherentist view, it is false that reasons form a linear structure that terminates with a basic belief.

    Cling finally argues that my conservative position represents a kind of arbitrariness to belief that makes it hard to distinguish from skepticism. Cling argues thusly. On my view, a person has some right to maintain any belief whatsoever when held in the special state of empty evidence. But, while it may be practically valuable to maintain such beliefs, it is difficult to see what epistemic value those beliefs have. Epistemic value requires reasons to think that those beliefs are true and my conservative position explicitly acknowledges that in such a state a person lacks such a reason.

    I agree with Cling that the epistemic value attaching to knowledge requires reasons. On my view, a subject has reasons for her beliefs to the extent that those beliefs are parts of a virtuous explanatory system that beats relevant competitors. Explanatory coherence is a significant epistemic achievement. While beliefs that have only conservative justification are held without reasons, beliefs that are able to be integrated into a virtuous explanatory system have the best kind of reasons for them. They fit together so that one’s perspective on the world is a unified whole that answers many challenges. Thus, as I see it, the sense in which a belief may be “arbitrary” is the sense in which a belief can be a starting point on the long road of inquiry aimed at knowledge.