When epistemologists argue against coherentism, they often do so by focusing on issues that are at best ancillary to the heart of coherentism.  Here I focus on two such criticisms.  The first is that some argue against coherentism by suggesting that it fails to provide a role for experiential input into justification.  The second is coherence is understood holitically, coherentists have no resources to explain the difference between a properly and an improperly based belief.  I argue that these apparently different criticisms are actually, at bottom, the same, and that they are mistaken because they mistake the heart of coherentism for particular versions of the view.


Coherentists' Distractions

 

The heart of coherentism is found in two aspects, one negative and one positive.  On the negative side, coherentism is a contrary of foundationalism, the view that the epistemic status of our beliefs ultimately traces to, or derives from, basic beliefs.  The positive aspect is found in the claim that the relation of coherence itself is responsible for any positive epistemic status possessed by beliefs. 

Unfortunately, when epistemologists argue against coherentism, they often do so by focusing on issues that are at best ancillary.  For example, some argue against coherentism by suggesting that it fails to provide a role for experiential input into justification.  If such criticisms are telling, they are telling only given the assumption that coherence is a relation solely among beliefs.  Coherentists have agreed that it is, but if we focus only on the heart of the position described above, it is not immediately obvious why they have done so.

Other times, an apparently unrelated difficulty is highlighted.  It is often thought that coherentism is at its best when it comes in holistic form--that is, when it maintains that it is the entire system of beliefs that conveys positive epistemic status on any particular belief.  The difficulty is that if the epistemic support relation is entirely systemic in this way, coherentists have no resources to explain the difference between a properly and an improperly based belief.[i]  On the usual account of the distinction, a properly based belief is one that is held because of, or on the basis of, the evidence for it, but if evidence is always systemic, then a properly based belief will have to be based on the entire system of beliefs, and that is psychologically implausible.


This worry is sometimes carried to an extreme.  Suppose one thinks that holistic coherentism is not only the most attractive version of the theory, but because of the miserable implausibility of any hint of non-holistic coherentism, the holistic variety is all there is.[ii]  If the above criticism is correct, then coherentists may find themselves in the position of having to maintain that all warranted beliefs are properly basic.  For if holistic coherentists cannot draw a distinction between properly and improperly based beliefs, every belief will have automatically survived all requisite tests for warrant just by cohering with the relevant system.  If a belief is properly based when it has survived all appropriate scrutiny, then all warranted beliefs will be properly basic, according to coherentism.

Neither criticism strikes at the heart of coherentism; at least, not obviously.   Because of this, one might wonder whether coherentism as such is threatened by these criticisms, or whether the perceived difficulties are mere distractions for coherentists.  I will argue the latter.  The above problems are mere coherentists' distractions because they focus attention elsewhere than on the important tasks for coherentists, which are to clarify the nature of coherence and demonstrate the superiority of coherentism to foundationalism.

More important and surprising, in spite of the fact that the above criticisms of coherentism appear unrelated, they are not.  They make precisely the same mistake in appraising the credentials of coherentism.  This mistake traces to the importance of the regress argument in the history of epistemology.  In that argument, one of the premises is that any non-basic warranted belief must obtain its warrant by some relation to other warranted beliefs.  This point leads to the assumption that the relations of  warrant impartation is the fundamental relation of interest to epistemological theory.  That is, one effect of the centrality of the regress argument to the history of epistemology is a myopic focus on the warrant impartation relation.  Whether the imparting of warrant is conceived of defeasibly or indefeasibly, the focus remains the same:  the question of positive epistemic status for non-basic beliefs turns on questions about the warrant impartation relation between a particular belief and other warranted beliefs.


My counterargument is that there are many more relations here than epistemologists have imagined.  That is, there are relations between beliefs and between experiences and beliefs that are relevant to the epistemic status of non-basic beliefs, but which are not themselves relations of warrant impartation.  A simple case arises from the character of defeasible reasoning itself.  Suppose believing p defeasibly warrants believing q; that is, suppose p and q stand in the relation of warrant impartation from the former to the latter.  Suppose further, however, that the transfer of warrant in a particular case is defeated by some further belief in d.  In such a case, p&d do not impart warrant to q.  To complicate things further, however, suppose that there is a further claim o, which has the following properties.  First, it is an overrider of defeater d, so that p&d&o imparts warrant to q.  Second, suppose that o by itself imparts no warrant to q.  If so, we have a very simple case of the sort on which I want to focus.  We have a case in which some claim can be relevant to the warrant of a non-basic belief, but not in virtue of standing in the warrant transfer or warrant impartation relation to that belief.  It is, of course, a part of a larger condition which itself stands in that relation, but the point remains nonetheless that a claim by itself can be relevant to positive epistemic status without standing in a warrant transfer relation to that which it helps warrant.

This point is a simple one, but it is one overlooked by both criticisms of coherentism above.  The objections arising for coherentism concerning the role of experience in justification and the importance of proper basing both founder on the possibility of relations that do not involve warrant impartation but which are nonetheless relevant to positive epistemic status.  One such relation is discussed above, but there is a more fecund example of such that is relevant here.  The fecundity of this relation will appear as we see how it undermines both of the above criticisms of coherentism in one fell swoop. 


The relation I have in mind is J.L. Mackie's concept of INUS conditions, which he uses to explain causation.[iii]  An INUS condition of some event e is an Insufficient but Non-redundant condition of e which is part of a larger condition which is itself Unnecessary but Sufficient for e; according to Mackie, one event is a cause of another if and only if, roughly, the first is an INUS condition of the second.  For example, throwing a baseball at a window can cause it to break, but not because the throwing is sufficient in any strong sense for the window to break (an equal and opposite force from the other side of the pane would have resulted in no damage to the window).  Neither is the throwing necessary for the window to break:  the window could have been broken by a sonic boom, for example.  Yet, there is a larger condition, according to Mackie, including the absence of contrary forces, of which the throwing of the ball is a non-redundant component, which is sufficient for the window to break.  Hence throwing a baseball causes the window to break because throwing the ball is an INUS condition of the breaking of the window.

This relation can be put to use in epistemology.  Using it, we can distinguish between, for example, something's imparting positive epistemic status to a particular claim, and that same thing being an INUS condition for positive epistemic status.  This distinction, I will argue, allows holistic coherentists to clarify the distinction between basic and non-basic beliefs, and it also allows a coherentist to appeal to both beliefs and experiences in defining positive epistemic status.  The ultimate lesson is that expanding the vision of epistemologists regarding the variety of relations that are relevant to the epistemic status of belief can prevent misleading appearances concerning the demise of coherentism and rid coherentists of criticisms which do nothing more than distract coherentists from their proper tasks, which are to clarify the coherence relation and argue for the superiority of the resulting theory to other theories of justification or positive epistemic status.  To demonstrate these points, I begin with the claim that coherence is a relation among beliefs. 

 

Coherentism and the Doxastic Assumption


Coherentism is usually understood as a position that maintains that justification or positive epistemic status is exclusively a function of the relationship between beliefs.  Following John Pollock, I will call this assumption the Doxastic Assumption; he says this assumption implies "that the justifiability of a belief is a function exclusive­ly of what beliefs one holds--of one's "doxastic state"."[iv]  As Keith Lehrer puts it, "There is nothing other than one's belief to which one can appeal in the justifica­tion of belief.  There is no exit from the circle of one's beliefs."[v]  Pollock and Lehrer are not alone in their endorsement of the centrality of the Doxastic Assumption to coherentism, for both coherentists and non-coherentists have joined together in affirming this characterization. Yet, it is not obvious why there is such agreement.  Coherentists maintain that justification is a matter of coherence with an appropriate system of information, and one such type of system is one's doxastic system--that system of information encoded in one's beliefs.  There are other systems of information, however; systems of information composed of what one's culture accepts, systems of information involving claims that should have been accepted but were not, systems of information encoded in mental states of any form whatsoever.  Why should coherentism as such be forced into restricting systems of information to those encoded in the form of belief?

Consider Lehrer's reasons for affirming that claim, for example.  His argument runs as follows:

This [the claim that there is no exit from the circle of beliefs] might not seem obvious.  It might, for instance, seem that one can appeal directly to experience, or the testimony of others, to justify one's beliefs.  But this is illusory.  Sense experi­ence, whether commonly casual or carefully controlled, always leaves open the question of what we are to believe.  The prick of sense often elicits ready consent, but what we believe in the face of sensory stimulation depends on our antecedent convictions.  For example, imagine we believe we see something red before us, and this belief arises so naturally and quickly that no other belief seems to be involved.  But we are enmeshed in our beliefs.  We believe our circumstanc­es are those in which we may trust our senses and, consequently, that there is little chance of error.  If we believed instead that the chance of error was great, we would resist responding with such perceptual belief.  Thus the stimulation of the senses elicits belief through the mediation of a system of antecedent beliefs.[vi]


There are two points to note about this argument.  First, the argument fails.  The conclusion Lehrer intends to demonstrate is that justification is completely a matter of the relationship between beliefs, which he expresses in the epigram that there is no exit from the circle of beliefs.  Lehrer's argument, however,  demonstrates at most something weaker:  that justificatory status always depends on beliefs in some way or another.  He does not show that experience, for example, plays no role in justification; instead, he shows at most only that it cannot by itself account for the epistemic status of belief.  Lehrer fails to address the further claim that experience might nonetheless be necessary for justification, and if experience is necessary by yielding partial justification for such beliefs, the Doxastic Assumption is false.  Second, and more important, even if the argument were successful, it does not show that coherentism itself is committed to the Doxastic Assumption.  Instead, it would at most show that the only plausible versions were so committed.  So this argument fails to give us a reason for characterizing coherentism in terms of such a commitment. 

Perhaps Lehrer only intends to offer what he considers to be the best version of coherentism, and so shouldn't be accused of poor metatheory regarding the proper construal of coherentism.  Other coherentists are more metatheoretically explicit, however.  Donald Davidson, for example, claims that the Doxastic Assumption is distinctive of coherentism.  He claims that coherentism is the view that nothing but beliefs can be appealed to in an account of the justification of our view of the world, and he argues that nothing can count as a reason for holding a belief except another belief.[vii]  He argues against any attempt "to ground belief in one way or another on the testimony of the senses,"[viii]  ­­­and he summarizes the case against such a view as follows: 

The relation between a sensation and a belief cannot be logical, since sensations are not beliefs or other propositional attitudes.  What then is the relation?  The answer is, I think, obvious:  the relation is causal.  Sensations cause some beliefs and in this sense are the basis or ground of those beliefs.  But a causal explanation of a belief does not show how or why the belief is justified.[ix]


This argument and the characterization of coherentism which it accompanies should be rejected, however.  First, coherentists need not limit their account of justification to relations between actual beliefs.  Consider Lehrer's verific knower version of coherentism.[x]  A simplified version of it implies that a belief is justified just when the person holding the belief would regard it as the best way to achieve the goal of getting to the truth and avoiding error, were that person purged of all beliefs held for motives other than a disinterested concern for the truth.  A simple emendation of this theory leaves its coherentism untouched but nonetheless fails to define coherence over actual beliefs.  On that emendation, not only do we imagine a person purged of beliefs of untoward motivation, we also imagine the person with the additional beliefs he or she would have had, had they been motivated in an epistemically more propitious manner.  On such a theory, coherence is not defined over actual beliefs, but the theory is coherentist nonetheless because it understands justification in terms of coherence with some preferred system of information.

In the above case, the preferred system is encoded in (counterfactual) beliefs, but that form is not obviously essential to the coherentist structure of the theory.  For, contrary to the possible states Davidson considers, there are a variety of non-doxastic states.  One such state is what Davidson calls a sensation state, a state that plays a causal role in belief but is not itself a propositional attitude.  There are other states to consider however.  One such state is an appearance state, which is a propositional attitude.  It can appear to one that, e.g., one's spouse is angry, and the content of the 'that' clause here demonstrates the propriety of labelling it a propositional attitude.  Furthermore, some such states surely fall under the rubric 'the testimony of the senses', and is not to be confused with any belief state that has the same propositional content.  One's appearances can be known to be misleading on certain occasions, so that one fails to believe that p even though it appears to one that p.  So Davidson's argument fails in virtue of a procrustean appraisal of the ways in which the testimony of the senses might be employed in a theory.  Moreover, there is no prima facie implausibility in the idea that a certain belief can cohere with a system of information, where some of that system is encoded in the form of appearance states instead of belief states (or counterfactual belief states). 


It is a bit puzzling that, though Davidson defines coherentism in terms of the Doxastic Assumption, he never explains why he does so.  Perhaps he is following tradition here, for such an understanding is nearly universally shared, or at least unopposed.  Yet, agreement in epistemological metatheory no more makes for truth than it does elsewhere.  The consensus opinion might be no more than a result of undue focus on the kinds of coherence theory typically offered in opposition to extant versions of foundationalism.  So, if coherence must be defined relative to a doxastic system, we are yet to see a good reason for thinking so. 

I submit that the argument presupposed by this understanding of coherentism is that, without an identification of coherentism with the Doxastic Assumption, coherentism collapses into a version of foundationalism.  In particular, coherentists need to oppose the regress argument for foundationalism, according to which justification must trace back to basic beliefs.   To be a foundationalist, then, requires at a minimum that one adopt an asymmetrical picture of justification.  On it, some beliefs are justified by being based on other beliefs, but there is also a class of beliefs that are justified in some other way.

So one way to avoid foundationalism is to offer a theory that denies asymmetry altogether, and affirming the Doxastic Assumption takes us part of the way at least to such a goal.  Even so, denying the asymmetry feature is not necessary for avoiding foundationalism.  For there are well-known examples of coherence theories that involve an asymmetry feature.  Explanatory coherence theorists, for example, sometimes claim that the explainers are justified differently from the way in which the explainees are justified.[xi]  Such coherence theories shows that asymmetry alone is not sufficient to guarantee that a theory is a foundationalist one. 


In addition to asymmetry, foundationalists also insist on some type of self-warrant or intrinsic warrant for foundational beliefs.  Minimally, relations to other beliefs cannot be the entire story, and a typical way to avoid having relations to other beliefs be the entire story is to claim that experience itself plays a role in virtue of imparting some degree of positive epistemic status to foundational beliefs.  Not even explanatory coherence theories can stomach such a role for experience, for it is essential to coherentism to oppose the foundationalist idea that some degree of warrant can be imparted to beliefs directly by experience.  If so, however, it looks as if any denial of the Doxastic Assumption that allows experience to play a role in justification will make one a foundationalist. 

This worry that leads to the marriage of coherentism and the Doxastic Assumption is unfounded.  The argument would work if there were only one relation that could hold between sensory experience and the epistemic status of beliefs, the relation whereby the former imparts some degree of epistemic status to beliefs.  There are other options, however.  One such option appeals to the concept of INUS conditions discussed earlier.  Perhaps experience imparts no degree of positive epistemic status to belief, but is rather an INUS condition for any belief to have positive epistemic status.  On such a picture of the relationship between experience and belief, some beliefs would depend for their epistemic status on experience, but not in a foundationalist way.  For experience would be Insufficient for any degree of positive epistemic status; it would simply be false that any belief receives any degree of warrant from experience.  Experience would nonetheless be a Non-redundant component of a larger complex which was itself Unnecessary but Sufficent for the belief to have positive epistemic status.  On such a picture, experience would play an important part in the story about positive epistemic status, but it would not play the kind of role distinctive of foundationalism.  The resulting theory would have the justification of some beliefs depending on experience, but not in a way that yields any kind of intrinsic or self-warrant to those beliefs--not in a way, that is, that relies on any impartation of even the tiniest degree of warrant to beliefs from experience.  For experience is simply insufficient, on this view, for any degree of positive epistemic status.  In this way, coherentists can appeal to experience without endorsing foundationalism.


Moreover, if one considers the motivation for coherentism, there is no particular reason for insisting that the system of information, in relation to which a belief is warranted, must be a system of doxastic information.  What coherentists oppose is the foundationalists' idea that warrant rests on or arises out of experience directly; they believe, instead, that justification arises out of interdependence within a coherent system of information.  In order to avoid foundationalism here, one does not need to deny experience every kind of explanatory role whatsoever regarding warrant; instead, one need only deny the distinctively foundationalist claim that some degree of positive epistemic status is imparted to belief by sensory experience itself.  Doing so leaves a theorist free to assert that warrant is a matter of coherence with an appropriate system of information, thereby affirming coherentism without implying which propositional attitudes such information must be encoded in, if such information must be encoded at all.

Why should this point about the variety of forms in which coherentism can appear matter?  The reason it matters is that many of the criticisms aimed at coherentism have force only against versions of coherentism wedded to the Doxastic Assumption (if they have any force at all).  First, consider the input problem, which BonJour formulates as follows:

Coherence is purely a matter of the internal relations between the components of the belief system; it depends in no way on any sort of relation between the system of beliefs and anything external to that system.  Hence if, as a coherence theory claims, coherence is the sole basis for empirical justification, it follows that a system of empirical beliefs might be adequately justified, indeed might constitute empirical knowledge, in spite of being utterly out of contact witht he world that it purports to describe.  Nothing about any requirement of coherence dictates that a coherent system of beliefs need receive any sort of input from the world or be in any way causally influenced by the world.[xii]


The input problem concerns the relationship between a system of beliefs and the external world.  Whatever the force of the objection, however, it disappears as a problem unique to coherence theories once experience is allowed to play a role.  Suppose, for example, that coherence is defined over both beliefs and appearance states, with appearance states being INUS conditions for positive epistemic status.  Then coherentists have every bit as much right as do foundationalists to claim that input from the world plays a role in determining which beliefs are justified.  So the first benefit of properly grasping the nature of coherentism is that it disarms the attempt by foundationalists to undermine coherentism with the input problem. 

A second problem is the data problem.  Put metaphorically, the problem is how the entire system of justification gets "off the ground" in the first place.  To get off the ground, there needs to be something with which a belief must cohere, something in place prior to the formation of the belief itself.  Consider, for example, the early stages of learning.  Even such early stages involve the formation of justified beliefs.  Yet, coherentism as traditionally understood borders on having to deny such a claim, in virtue of the fact that there is little in place at such an early stage with which beliefs can cohere.  Coherence theories that deny any any asymmetry between the justification of some beliefs and the justification of others, and which also define justification in terms of coherence with the system of information encoded in one's actual beliefs, face such a problem.  For such early learning involves the development of a system for future beliefs to cohere with, and thus seems incapable of involving such coherence itself. 

Other coherence theories have no easier time with the data problem.  Explanatory coherence theories, for example, try to respond to the problem by suggesting that ­some beliefs are justified by explaining other beliefs while these latter beliefs are justified by being explained.  The difficulty here is obvious, as Lycan explains: 

­The present suggestion is that we begin with a set of initial data, explain these by making certain hypotheses, defend the hypotheses on the basis of their ability to explain the data, and then defend the initial data on the ground that they are explained by a set of justified explanatory hypotheses.  Right:  Any elementary logic student would spot the circularity.[xiii]


Lycan, like BonJour, prefers to take the data to be spontaneous beliefs.  For BonJour, observational beliefs are a subset of spontaneous beliefs, and the data problem is addressed by having observational beliefs function as the data to which a system of beliefs is responsible.  Such observational beliefs are nonetheless justified by their relation to other beliefs:  beliefs such as that the cognitively spontaneous belief in question is of kind K and arises in circumstance C, and K beliefs formed in C are most likely true.[xiv]  Lycan defends the justifiedness of spontaneous beliefs differently, by appealing to a conservatism doctrine:  the doctrine which claims that one should accept those spontaneous beliefs that seem to be true.[xv]

Neither of these responses to the data problem is wholly satisfactory.  First, a circularity threatens BonJour's account unless he can explain how the beliefs which underly the justification of any spontaneous belief do not depend on spontaneous beliefs of the same kind for their justification.  For example, if spontaneous beliefs of kind K formed in circumstance C are most likely true, how is this belief justified except on the basis of having formed many spontaneous beliefs of that kind in those circumstances and noting that most of them are true?  One doesn't get knowledge about the general reliability of perception without first having many perceptual beliefs; and the process of finding out that perception is reliable is one riddled with instances of justified beliefs.  Moreover, BonJour himself admits that his account of justification implies radical skepticism (in virtue of depending on what he calls "the Doxastic Presumption," the presumption that we do have an adequate grasp of which beliefs we hold),[xvi] and the reason it does infects the justification given above for observational beliefs, thereby erasing any comfort that might be found in his approach to the data problem.  For if the data problem can be solved only within a version of coherentism that embraces radical skepticism, such a "solution" is hardly worthy of the name.

Lycan's proposal avoids this problem, but faces another.  According to Lycan, one should accept those spontaneous beliefs that seem to be true, i.e., the data beliefs are those that are spontaneous and seem to be true.  One wonders, however, why spontaneity plays a role here at all.  Why not, for example, simply favor those beliefs of yours that seem to you to be true (are there some that don't)?  ­Such radical conservatism is much stronger than Lycan endorses, but it is hard to see what reason he has for avoiding it.  To find such a reason, he needs to tell us what it is about spontaneity as such that is justification-imparting:  why spontaneity itself is an epistemic­ally favorable characteristic of a belief. 


If, however, we deny the Doxastic Assumption by defining coherence over both appearances and beliefs, the data problem is not severe.  One way of specifying the class of data beliefs is in terms of those with a content also characterizing some appear­ance state.  The justification of such beliefs would depend on coherence not only with this particular appearance state, but also with other data beliefs and with any system already in place.   Such a view involves no circularity worries, nor does it employ any conservativism doctrine, though it does precisely what Lycan and BonJour seem to be trying to do:  get a class of data beliefs that are tied to experience. 

We may conclude then that there are significant advantages to distinguishing between the impartation of warrant relation and INUS conditions for warrant.  When it is assumed that only the former relation can obtain between belief and experience, one can avoid foundationalism only by declaring allegiance to the Doxastic Assumption.  If, however, one recognizes the latter possibility, coherentism is freed of criticisms that arise only when it is conjoined to that assumption.  The result is an understanding of coherentism whose credentials can be compared and contrasted to those of foundationalism without any unnecessary distraction arising from perceived problems with the Doxastic Assumption. 

 

Coherentism and the Basing Relation


The Doxastic Assumption is not the only distraction that arises in recent epistemology when coherentism's credentials are purportedly on review, however.  Another one has to do with the relationship between coherentism and the basing relation, the relation which distinguishes beliefs that have only a justified content from those beliefs where the believing itself is justified.  For example, Holmes and Watson may have precisely the same information about who committed a particular murder, and both may believe that the butler did it.  Holmes' belief may nonetheless have a virtue not possessed by Watson's.  For Watson may have acquired the belief by a sturdy knock on the head after falling down a flight of stairs, whereas Holmes has "deduced" the guilt of the butler from the information that both he and Watson possess.  In such a case, Watson's belief may have a content that is justified, but Holmes has something more.  Whereas Watson's believing is not justified even though what he believes is justified (by the information he and Holmes share), Holmes' believing is itself justified, for he holds the belief on the basis of, or because of, the evidence for it.  When warrant obtains for the content of a belief, I will say that it is propositionally warranted; when a belief is both propositionally warranted and properly based, I will say it is doxastically warranted.[xvii]

Recently, some have suggested that coherentism is committed to peculiar views about the basing relation, views that cast doubt on the adequacy of coherentism.  As we shall see, appeal to INUS conditions absolves coherentism of the charge and removes another distraction from an accurate assessment of the merits of coherentism.

The discussants I have in mind are Alvin Plantinga and John Pollock.  Plantinga characterizes coherentism so that it must deny that there are non-basic warranted beliefs,[xviii] and John Pollock thinks that an important kind of coherentism cannot explain the difference between a properly and an improperly based belief.[xix]  These two claims are connected.  If Pollock is right, and if Plantinga were to have warranted reservations about the possibility of a coherentism other than the one Pollock criticizes, Plantinga's construal of coherentism would be implied by Pollock's criticism.  To see this, consider Plantinga's characterization of coherentism:  

Current lore has it . . . that the coherentist does not object to circular reasoning at all, provided the circle is large enough...


But why saddle him with anything so miserably implausible?  There is a much more charitable way to construe his characteristic claim.  He should not be seen as endorsing circular reasoning. . .. His suggestion, instead, is that coherence is the sole source of warrant.  He is instead pointing to a condition under which a belief is properly basic. . ..  On his view, a belief B is properly basic for a person S if and only if B appropriately coheres with the rest of S's noetic structure. . .. (p. 77ff.)

Plantinga here proposes that coherentism is a special kind of foundationalism because it implies that there are properly basic beliefs.  Even worse, coherentists are foundationalist zealots--not only are some warranted beliefs properly basic, all of them are.

These claims are shocking.  Coherentists are wont to assert, not that all warranted beliefs are properly basic, but rather that none of them are.  Coherentists would react strongly to being characterized as special kinds of foundationalists, inclined perhaps to treat such pronouncements as projections by harried  foundationalists instead of insight into coherentism.  We might suspect, then, that Plantinga has unusual senses of the crucial terms employed in his characterization of coherentism.  The evidence, however, shows otherwise. 

For Plantinga, which basic beliefs are properly basic is a matter for each particular theory to clarify (p. 70), and  Plantinga endorses two principles about the basing relation itself: 

(1) If one belief is based on another, then the second is a cause of the first; and

(2) If one explicitly infers a belief from another, then one bases the first belief on the second. (p. 70)

One obvious implication of claim (2) is that no basic belief is inferential in character, for by definition basic beliefs are beliefs that are not based on other beliefs.  Since inferential beliefs are not basic beliefs, Plantinga implies that coherentists must deny that any warranted belief is ever explicitly inferred from other beliefs.  That is, coherentists deny the possibility of warranted inferential beliefs.


Recall that Plantinga's explicit motivation for his construal of coherentism was charity, wanting to avoid saddling the coherentist with the "miserably implausible" claim that circular reasoning can sometimes generate warrant.  Such charity will leave the coherentist wishing Plantinga had already given at the office.  No coherentist I know of accepts such a characterization, and if some do, they are foolish indeed.  For it is patently obvious that warranted beliefs can be inferred ones. 

Plantinga's mistake in characterizing coherentism is contained in the quote above.  First, he correctly points out that the coherentist's "suggestion . . . is that coherence is the sole source of warrant."  But in the very next sentence, he also claims that the coherentist "is . . . pointing to a condition under which a belief is properly basic."  Since his discussion of coherentism focuses on this second claim, Plantinga may think that the two claims are equivalent, or that the second follows from the first. 

No such connection exists.  The first remark (that coherence is the sole source of warrant) is a claim about propositional warrant, that kind of warrant that accrues to the content of what one believes (or to the content of a claim one does not believe).  According to a coherentist, only coherence with an appropriate system is capable of generating such warrant.  The concepts of basing and proper basing employed in the second claim above (that the coherentist is pointing to a condition under which a belief is properly basic) are quite another thing.  These concepts have to do with doxastic warrant, the warrant a belief has when its content is propositionally warranted and the belief is properly based.  So, Plantinga has made a mistake in characterizing coherentism, one that arises by failing to recognize the distinction between propositional and doxastic warrant.

Somewhat surprising, it is not clear that Plantinga believes what he writes anyway.  He remarks later on two types of coherentism: 

The pure coherentist holds that all warranted propositions in a noetic structure are basic in that structure; no warrant gets transmitted.  The impure coherentist holds that some propositions may get their warrant by virtue of being believed on the basis of others; but the ultimate source of the warrant in question is coherence.  Both accept the view that coherence is the only source of warrant; and this is the central coherentist claim. (pp. 79-80)


This quote stands in contrast to the earlier quote from Plantinga, for here we have specified a version of coherentism (impure coherentism) according to which some beliefs are warranted but not properly basic, because they get their warrant from other beliefs.  What makes the view coherentist is its insistence that the ultimate source of warrant is still coherence. 

Perhaps a charitable way to interpret Plantinga is to take his earlier remarks about coherentism as applying only to pure coherentism.  Then we could say that he characterizes coherentism so that coherence is the sole source of warrant, and that this claim, when unsullied by impurities, yields the pure view that all warranted beliefs are properly basic.

This interpretation removes the contradiction but leaves the mystery, for we still have a characterization of one kind of coherentism that no coherentist should accept and that is still susceptible to the response that such a characterization arises only by ignoring the distinction between propositional and doxastic warrant.  The mystery could be solved in a way that absolves Plantinga if coherentist purity undermines any appeal to this distinction in such a way that the only choice left to coherentists is to agree that all warranted beliefs are properly basic, despite what they might wish to say.  It is here that Pollock's criticisms of holistic coherentism are relevant. 

Pollock says that there is a distinction between what I call propositional and doxastic warrant, and that any correct theory must be able to give an account of it.  But, he claims, holistic positive coherence theories cannot explain the distinction, for such a distinction would require a coherentist to maintain that every belief is a partial cause of every other belief because of the holistic picture of warrant adopted by coherentists.(p. 81)  So holistic coherentism should be rejected. 


This criticism can support Plantinga's characterization of  coherentism in the following way.  If Pollock is right, some coherentists must hold that no belief ever fails to be warranted by failing to be properly based.  To claim otherwise requires holding the hopelessly implausible claim that a belief is based on absolutely everything in one's entire belief system.  Recall that Plantinga, unlike Pollock, thinks the linear view according to which reasoning in a large enough circle is acceptable misinterprets the claims of coherentism.  In the passage quoted earlier, he notes that he sees no good reason to interpret the coherentist's characteristic claim in this "miserably implausible" fashion.  This remark suggests that Plantinga thinks there is no such thing as a linear, or non-holistic, version of coherentism.  If we were to conclude, then, that coherentism can only come in the form Pollock criticizes, and that criticism is correct in claiming that the coherentist cannot give an account of proper basing, we get the following.  The coherentist must maintain that no belief ever falls into epistemic disfavor by being improperly based, for to fall into disfavor on such grounds would require a theory of the basing relation that could distinguish properly from improperly based beliefs.  And if the coherentist must maintain that no belief ever falls into epistemic disfavor on grounds of improper basing, the coherentist must admit that every belief is automatically based properly, i.e., that it passes any legitimate basing test for (doxastic) warrant.  So we get the Plantingian characterization if Pollock is right:  for a true coherentist, every warranted belief is properly basic.

The important question regarding the difficulties which arise for coherentism in these related discussions is whether Pollock's criticism of (one kind of) coherentism can be sustained.  In a word, it cannot.  Pollock assumes that the coherentist must clarify the concept of proper basing in terms of a causal  relation between that which propositionally warrants a belief and the belief in question.  This assumption involves two requirements.  The first is that the basing relation is a species of causal relation; I will grant that point in what follows.  The second concerns what the relata of the basing relation; what the basing relation is a relation between.  According to Pollock, it must be a relation between that which propositionally warrants a belief and the belief  itself.  I will focus on this second claim. 


There is no good reason for a coherentist to accept such a claim.  Propositional warrant may be as systemic an affair as you please, and yet doxastic warrant depend nonetheless on some special components of the system.  Consider again INUS conditions for warrant, for example.  Such a condition might be a cause of belief, and there is no reason a coherentist cannot appeal to such conditions in an account of proper basing.  Such an appeal amounts to a denial of the claim that the basing relation is a relation between that which propositionally warrants a belief and the belief itself; according to the holistic coherentist, that relation obtains only between the entire relevant system and the belief itself.  Nonetheless, the entire relevant system might contain INUS conditions for the warrant of the belief in question, and the coherentist can claim that in order for the belief to be doxastically warranted, it must be based on some INUS condition or conditions for the warrant of that belief.  

Some sketchy examples of what such a theory might look like may prove helpful.  Consider subjective versions of coherentism.  Such versions can maintain that one's system of beliefs contains a (subjective) theory of epistemic relevance that places constraints on appropriate basing.  On such a theory, having an appropriately based belief (i.e., a belief that is not disbarred from candidacy for doxastic warrant because of some defect with regard to how it is based) might require being aware of just which elements in the system count as epistemically relevant to the belief (where INUS conditions for warrant count as epistemically relevant conditions even though they impart no warrant).  Alternatively, a theory of this sort might require that the explanation (the best one and perhaps the one the person in question accepts or would accept on reflection) of why one holds the belief conform to that subjective theory of epistemic relevance.


Imagine, for another example, a Bayesian account of warrant in terms of degrees of belief.  The version I imagine employs diachronic dutch books as a constraint on warranted degrees of belief, so that one's warranted degree of belief tomorrow is a function of one's conditional degrees of belief today, conditional on what future experience might teach.[xx]  Such a theory is fully holistic because warrant obtains on such a theory only when the entire set of beliefs is probabilistically coherent.  What is important about the view in the present context is that one's conditional probabilities today contain an implicit theory of epistemic relevance, a theory that implies that some new information is relevant to some degrees of belief and not others.  By containing such an implicit theory of relevance, a Bayesian account of warrant can add a basing requirement without adopting the absurd viewpoint that the entire system of (degrees of) belief must be causally responsible for every properly based degree of belief.  It can do so by insisting that its theory of epistemic relevance specifies only INUS conditions for warrant. 

Pollock might wish to classify these theories as non-holistic, but his account of that distinction fails to yield that result.  He says that a linear coherence theory "embraces essentially the same view of reasons and reasoning as a foundations theory," one according to which "P is a reason for S to believe Q by virtue of some relation holding specifically between P and Q.  A reason for a belief is not automatically the set of all one's beliefs."  He says a holistic coherence theory claims that, "in order for S to have reason for believing P, there must be a relationship between P and the set of all of his beliefs (where this relationship cannot be decomposed into simple reason relationships between individual beliefs)." (p. 73) 


This distinction fails to show that the above theories are non-holistic, for an INUS condition for warrant for p is simply not a reason, defeasible or otherwise, for p.   A reason for p is something that imparts (perhaps defeasible) warrant to p, and an INUS condition for warrant is Insufficient for the impartation of any warrant whatsoever.  Thus, a coherentist can affirm a holistic view of reasons, where only systemic relations count in favor of the warrantedness or positive epistemic status of belief, and can nonetheless isolate INUS conditions of positive epistemic status with reference to which one defines proper basing.  Consider the Bayesian view above again.  One's probability for p given q constrains one's opinion about p upon learning q in part because the condition cited is not simply q, but rather q plus all of one's background information.  So what is sufficient for the warrant p acquires when q is learned is some larger condition (q plus background information) of which q is a non-redundant component (because the background information alone does not warrant p).  Further, the larger condition, though sufficient for the warrant of p, is not itself necessary.  So the Bayesian view is already three-fourths of the way toward q being an INUS condition for warrant with respect to p; all that needs to be affirmed is that q on its own never imparts any degree of warrant.  Certainly, that option is open to Bayesians, and if it is taken, the conditional probabilities of today that constrain future opinion do so by specifying INUS conditions for warrant.  By taking this option, the Bayesian can avoid the charge of holding a linear coherence theory and still use INUS conditions for warrant in order to define proper basing. 

Such a theory is a holistic one on Pollock's category scheme, for the concept of a reason for belief (i.e., that which imparts warrant to belief) is entirely systemic (what imparts warrant is what you learn together with your entire system of background information), not capable of being decomposed into reason, i.e., warrant-imparting, relationships between individual beliefs. 

Because of their holistic view of reasons, holistic theories such as the Bayesian and subjective coherence theories above retain the advantages a holistic theory has in addressing the regress argument for foundationalism, for that argument presupposes a linear view of warrant transmission.  Of course, a foundationalist could try to resurrect something like the regress argument that applies to INUS conditions for warrant, but the prospects for such an argument are not good.  It may be true that every belief that is an INUS condition for warrant will itself need to be warranted, but holistic theories are at no obvious disadvantage in accounting for such warrant.  Even if answering such questions leads the coherentist back to beliefs that are warranted by a condition that includes appearance states as INUS conditions, no obvious disadvantage accrues to the view defended.  All that would follow is that such a coherentist posited an asymmetry between the justification of some beliefs and the justification of other beliefs.  As we have seen, however, such an asymmetry alone does not imply foundationalism.  If the coherentist were to assert that experience is not just an INUS condition for warrant, but imparted some degree of warrant directly to beliefs, there would be a problem.  But the coherentist should not say that, and in the absence of such a claim, no foundationalist implications are to be found.[xxi]


One might worry that a theory of proper basing which appeals to INUS conditions will not work on grounds that everything in the relevant system will be such an INUS condition.  Such a worry can be allayed easily.  I'd have the same warrant for thinking that I exist even if I didn't believe my grandmother is quirky, or even if I believed the opposite.  So the latter belief plays no important explanatory role regarding the warrant of the former belief, and hence is not a Non-redundant part of the belief system.  So not every element of a system of information is an INUS condition of warrant for every (other) belief.

One might still worry that too much of a belief system will be INUS conditions, and hence that the coherentist will have to hold that a belief will have to be based on a substantial part of the system.  This concern highlights the need for something I lack:  a good theory of the basing relation.  The force of the criticism can nevertheless be blocked, for the difficulty is not unique to coherentism.  The same problem arises for a defeasible reasons theory.  As we have seen, on such a theory there can be multiple reasons for believing a particular claim:  p can be a reason to believe q, be defeated by d which is in turn overridden by o, yielding that p&d&o is also a reason to believe q.  Furthermore, the heirarchy of defeaters and overriders is potentially unlimited, yielding the result that there can be a potentially unlimited number of reasons for believing any particular claim.  The question for such theories is, which reasons should the belief be based on?  This question raises precisely the worry faced by the coherentist that too much of the system will be INUS conditions for warrant.  Furthermore, both kinds of theorists will answer the worry in one of two ways:  either isolate some of the INUS conditions or defeasible reasons as privileged when it comes to basing, or argue that there is nothing especially implausible about insisting that one base one's belief on the entire collection of INUS conditions or defeasible reasons (for, as all should recognize, belief formation is a very complex thing and it should not surprise us if very much of our belief system is causally responsible in one way or another for belief).  We can thus legitimately ignore the worry that too much of the system will be INUS conditions for warrant, for the difficulty posed is not unique to coherentism and the coherentist can approach the problem in much the same way as other theorists.


So the point stands that in the subjective and Bayesian examples above, accounts of the basing relation are possible that are not psychologically implausible.  The adequacy of any such theory is, of course, another question.  But the mere possibility of formulating them shows that Pollock is mistaken.

With the failure of Pollock's argument goes Plantinga's construal of coherentism, for the only hope for that construal rested on denying that coherentists can help themselves to the distinction between doxastic and propositional warrant.  So Plantinga has misconstrued coherentism and Pollock has misapprehended its weaknesses.  Coherentism is the view that limits the sources of propositional warrant to coherence itself, but this (admittedly unenlightening) characterization leaves coherentists free to impose basing requirements of their choice on doxastic warrant.  Contrary to Pollock, coherentists can distinguish between propositional and doxastic warrant, and because they can, contrary to Plantinga, coherentism does not imply that any warranted beliefs are properly basic.

 

Conclusion


So we can see that the distinction between a warrant-imparting relation and an INUS condition relation is particularly fecund one in epistemology.  In particular, the distinction removes some important contemporary distractions from a fruitful appraisal of the merits of coherentism.  The heart of coherentism is found in its rejection of foundationalism and in its affirmation that justification, or positive epistemic status, is to be understood in terms of coherence.  My goal here is to focus attention on these fundamental claims of coherentism by removing tangential concerns.  Coherentism proper is not committed to the Doxastic Assumption, and faces no difficulty in holding that some beliefs are non-basic or in clarifying the distinction between properly and improperly based beliefs.  In each case, treating either appearances or individual beliefs as INUS conditions for positive epistemic status shows that coherentists can appeal directly to experience in its formulation and can adequately explain which beliefs in a system are properly based and which are not.  With this ground-clearing completed, one can look again at coherentism, in a clear light, undistracted by irrelevancies.[xxii]

 

Jonathan L. Kvanvig

Texas A&M University

 

‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑

Notes 



[i]John Pollock, Contemporary Theories of Knowledge, (Totowa, New Jersey, 1986).  Page references in the text to Pollock's view are to this work.

[ii]Alvin Plantinga, Warrant:  The Current Debate, (Oxford, 1993).  Page references in the text to Plantinga's views are to this work.

[iii]J.L. Mackie, The Cement of the Universe:  A Study of Causation, (Oxford, 1974).

[iv]John Pollock, Contemporary Theories of Knowledge, (Totowa, 1986), p. 19.

[v]Keith Lehrer, Knowledge, (Oxford, 1974), p. 188.

[vi]ibid.

[vii]Donald Davidson, "A Coherence Theory of Truth and Knowledge," pp. 307-319 in Ernest LePore, ed., Truth and Interpretation, (New York:  Basil Blackwell), 1986.

[viii]ibid., p. 310.

[ix]ibid., p. 311.

[x]Keith Lehrer, Knowledge.

[xi]See Keith Lehrer, Knowledge, for an assessment of the prospects of such a theory.

[xii]BonJour, p. 108.

[xiii]Lycan, p. 165.

 

[xiv]BonJour, pp. 117-118.

[xv]Lycan, p. 165-66.

[xvi]BonJour, p. 105.

[xvii]Here I ignore differences between the epistemological concepts of rationality, justification, and warrant.  There may be differences between these concepts that are important in some contexts, but the differences are irrelevant here.

[xviii]Alvin Plantinga, Warrant:  The Current Debate, (Oxford, 1993).  Page references in the text to Plantinga's views are to this work.

[xix]John Pollock, Contemporary Theories of Knowledge, (Totowa, New Jersey, 1986).  Page references in the text to Pollock's view are to this work.

[xx]For explication and discussion of such versions of Bayesianism, see Bas van Fraassen, Laws and Symmetry, (Oxford, 1989), especially Part II.

[xxi]For more on the need for self or intrinsic warrant to define foundationalism, and its relationship to INUS conditions of warrant, see "Can A Coherence Theory Appeal to Appearance States?", especially pp. 197-200 and 211-212.

[xxii]For comments on earlier drafts of this paper, my thanks to Michael Hand, Hugh McCann, and Mark Migotti.