Bonjour Coherence

Coherence theory of knowledge and justification LAURENCE BONJOUR

2

LAURENCE BONJOUR

Coherence theory of knowledge and justification1 Coherence theories of justification represent one main alternative to foundationalist theories of justification. If, as has usually been thought, possessing epistemic justification is one necessary condition (along with truth and perhaps others) for a belief to constitute knowledge, then a coherence theory of justification would also provide the basis for a coherence theory of knowledge. While some proponents of coherence theories have restricted the scope of the theory to empirical justification, others have applied it to all varieties of epistemic justification. (There are also coherence theories of meaning and of truth, as well as coherence theories of ethical or moral justification.) The initial contrast between coherence theories and foundationalist theories arises in the context of the epistemic regress problem. It is obvious that the justification of some beliefs derives from their inferential relations to other, putatively justified beliefs, and that the justification of these other beliefs may depend on inferential relations to still further beliefs, and so on, so that a potential regress of epistemic justification looms, with scepticism as the threatened outcome. The foundationalist solution to this problem is that one arrives sooner or later at basic or foundational beliefs: beliefs that are epistemically justified, but whose justification does not derive from inferential relations to any further beliefs and so brings the regress to an end. The defining tenet of a coherence theory of justification is the rejection of this foundationalist solution, the coherentist insisting that any belief (of the kinds to which the theory is applied) depends for its justification on inferential relations to other beliefs and eventually to the overall system of beliefs held by the believer in question. According to the coherentist, the justification of this system of beliefs is logically prior to that of its component beliefs and derives ultimately from the coherence of the system, where coherence is a matter of how tightly unified or interconnected the system is by virtue of inferential connections (including explanatory connections) between its members. Contrary to what this might seem to suggest, coherence theories do not deny that sensory observation or perception plays an important role in justification. What they deny is that this role should be construed in a foundationalist way, insisting instead that the justification of observational beliefs ultimately derives also from considerations of coherence. Specific coherence theories may also add other requirements for justification, thereby departing from a pure coherentism, while still avoiding foundationalism. While the idea of a coherence theory has often played the role of a dialectical foil, developed theories of this kind are relatively rare and are 1

BONJOUR, LAURENCE (1998). Knowledge and justification, coherence theory of. In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. Retrieved May 30, 2006, from http://www.rep.routledge.com/article/P009.

Coherence theory of knowledge and justification

3

often in serious disagreement among themselves. In this way, coherentism is much less a unified view with standard, generally accepted features, than is foundationalism.

1. History2 In contrast to foundationalism, the coherence theory of justification is a relatively recent innovation in the history of philosophy. Although it is possible, albeit with some strain, to construe Spinoza and Kant as advocating versions of coherentism, the first relatively clear-cut coherentist positions are those of the late nineteenth and early twentieth century absolute idealists, especially F.H. Bradley (1914) and Bernard Bosanquet (1920) (see Bradley, F.H.; Bosanquet, B.). Unfortunately, however, the views of these philosophers are marked by a pervasive failure to distinguish epistemological and metaphysical issues, making it hard to separate their coherence theories of justification from their distinct, though related, advocacy of coherence theories of the nature of truth. (A more recent version of essentially the same position, in which this distinction is clearly drawn, is found in the work of Brand Blanshard (1939).) Coherentism was also advocated in the 1930s by some of the logical positivists, mainly Otto Neurath (1933) and Carl Hempel (1934), in response to the foundationalist views of Moritz Schlick (1934) (see Neurath, O.; Hempel, C.G.; Schlick, F.A.M.). Neurath identifies coherence with mere logical consistency; he also, while retaining something like a justificatory appeal to observation, in effect identifies observational beliefs solely by reference to their content. He thus has no apparent response to what is perhaps the most central and obvious objection to coherence theories: that there will always be indefinitely many different coherent systems between which a coherence theory will provide no basis for a reasoned choice (see further below). Hempel avoids this problem to some extent by simply identifying observational beliefs as those beliefs with the right sort of content that are accepted by ‘the scientists of our culture circle’, but is able to offer no real rationale for such an identification. He also, like the idealists, fails to distinguish in any clear way between a coherence theory of justification and a coherentist account of the nature of truth. More recent coherentist positions, in contrast, generally repudiate the coherence theory of truth entirely and are more explicitly and narrowly epistemological in their character and motivation. The main arguments offered in their favour almost always derive from perceived objections to foundationalism, perhaps the central one being the charge that the foun2

BONJOUR, LAURENCE (1998). Knowledge and justification, coherence theory of. In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. Retrieved May 30, 2006, from http://www.rep.routledge.com/article/P009SECT1.

4

LAURENCE BONJOUR

dationalist can account for the status of the allegedly basic or foundational beliefs as genuinely justified (in the sense of there being some reason or basis for thinking them to be true) only by appealing to justificatory premises of some sort and so destroying the status of such beliefs as foundational (see Foundationalism). Thus coherentists insist that there is no way to appeal for justification to anything outside of one’s system of beliefs because any such supposed source of justification would have to be apprehended by the person in question in a belief or belief-like state before it could play any justificatory role, and then it would be the belief rather than the external item that was the immediate source of justification. As this suggests, coherentist positions are virtually always internalist rather than externalist in character, in that they insist that the basis for epistemic justification must be cognitively accessible to the believer in question; while an externalist version of coherentism is theoretically possible, it would have little philosophical point, since a foundationalist view would be vastly more straightforward if externalism were otherwise acceptable (see Internalism and externalism in epistemology). These recent coherentist views differ from each other in a wide variety of ways, and often seem to have little in common beyond their rejection of foundationalism and their invocation in some fashion of the idea of coherence (and indeed there is often room for doubt in a particular case about how thoroughgoing the former of these two aspects really is). Coherentism is one ingredient, though never developed in a fully systematic way, in the comprehensive and difficult philosophical system of Wilfrid Sellars (1963) (see Sellars, W.S.). The epistemological position of W.V. Quine is also frequently described as coherentist in character, though other features of Quine’s position, especially his claim that epistemology should be naturalized (reduced to psychology), make it difficult to decide whether his view is genuinely a version of coherentism, as opposed to a qualified version of foundationalism (see Naturalized epistemology). More overtly coherentist positions have been advocated by Gilbert Harman (1973) (influenced especially by Quine), Nicholas Rescher (1973, 1974, 1977), Keith Lehrer (1974, 1990), and Laurence BonJour (1985) (influenced especially by Sellars). As the foregoing suggests, coherence theories first arise as dialectical alternatives to foundationalism, rather than as views that are claimed to be initially plausible on their own. Their defence and elaboration must confront a number of standard problems and objections, with which any such view must seemingly deal in some fashion, and it is around these that the present entry is organized.

Coherence theory of knowledge and justification

5

2. The regress problem and non-linear justification3 The first standard problem arises from the epistemic regress problem itself. If foundationalism is repudiated (and if a genuinely infinite regress of justification is also rejected as psychologically impossible and in any case tantamount to scepticism), then the only remaining possibility for the outcome of the initial regress of epistemic justification seems to be a circle in which the chains of justification eventually loop back upon themselves. Incautious advocates of coherentism have sometimes seemed to endorse the idea that such a result is acceptable if only the circles are ‘large enough’. But the obvious objection to circular chains of justification, to which the size of the circle seems irrelevant, is that they involve circular reasoning and hence have no genuine justificatory force. This is essentially the reason that foundationalists give for rejecting the coherentist alternative and taking the regress problem to constitute a decisive argument for foundationalism. Perhaps the most standard coherentist response to this issue, stemming originally from Bosanquet (1920), is to reject the idea, implicit in most presentations of the regress problem, that relations of justification must involve a linear, asymmetrical order of dependence among the beliefs in question. They insist instead that justification, when properly understood, is ultimately holistic and non-linear in character, with all of the beliefs in the system standing in relations of mutual support, but none being epistemically prior to the others. In this way, it is alleged, any true circularity is avoided. Such a view amounts to making the system itself the primary unit of justification, with its component beliefs being justified only derivatively, by virtue of their membership in an appropriate sort of system. And the property of the system, in virtue of which it is justified, is of course specified as coherence.

3. The concept of coherence4 But what exactly is coherence? A second obvious problem for a coherence theory is to explicate and clarify this central concept. Intuitively, coherence is a matter of how the beliefs in a system of beliefs fit together or dovetail with each other, so as to constitute one unified, organized, and tightly structured whole. And it is clear that this fitting together depends on a wide variety of logical, inferential and explanatory relations among the components of the system. But spelling out the details of this idea, particularly in a way that would allow unproblematic assessments of comparative coherence, 3 BONJOUR, LAURENCE (1998). Knowledge and justification, coherence In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. May 30, 2006, from http://www.rep.routledge.com/article/P009SECT2. 4 BONJOUR, LAURENCE (1998). Knowledge and justification, coherence In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. May 30, 2006, from http://www.rep.routledge.com/article/P009SECT3.

theory of. Retrieved theory of. Retrieved

6

LAURENCE BONJOUR

turns out to be extremely difficult, in part at least because of its obvious dependence on more specific and still unsettled topics, such as induction, confirmation, probability and explanation. The strongest and most demanding conception of coherence, advocated by the idealists, specifies a coherent system of beliefs as one in which each member entails and is entailed by all of the others. It seems clear, however, that this strong conception is both unrealizable by any actual system of beliefs imaginable and also of dubious cognitive value, since it would seem to make all of the beliefs but one redundant and dispensable. (These problems may be mitigated somewhat, though certainly not eliminated, by remembering that the idealists have a quite broad conception of entailment, one in which, for example, relations of nomological necessity are regarded as a kind of entailment). At the opposite extreme, it seems equally mistaken to follow Schlick (1934) and some others in identifying coherence with mere logical consistency, since the beliefs of a logically consistent system might be entirely unrelated to each other, thus yielding no real degree of mutual support and no apparent reason for thinking that any of them are true. Somewhat more surprisingly, it also seems to be a mistake to make complete logical consistency even an absolutely necessary condition for any degree of coherence, as many coherentists have done. In light of such things as the preface paradox and general human fallibility, this would probably mean that few if any actual systems of belief are coherent to any degree at all, a result that seems unacceptably paradoxical (see Paradoxes, epistemic). If there is a tenable conception of coherence along these general lines, it must seemingly fall somewhere between the two extremes just discussed. Coherence will be a matter of degree, with logical consistency being a highly relevant but not absolute criterion. Coherence will also require a high degree of inferential interconnectedness in the system, involving relations of necessitation, both strictly logical and otherwise, together with probabilistic connections of various kinds. An important aspect of this is what might be called probabilistic consistency, that is, the absence of relations between beliefs in the system in virtue of which some are highly unlikely to be true in relation to others. A further important ingredient of coherence that is much emphasized in recent discussions is the presence of explanatory relations among the components of the system, thus reducing the degree to which the beliefs of the system portray unexplained anomalies. (If ‘inference to the best explanation’ is accepted as one species of inference, then such explanatory relations can be viewed as a kind of inferential relation – see Inference to the best explanation.) Indeed, some positions such as that of Harman (1973), and perhaps also Sellars (1963), go so far as to virtually identify coherence with the presence of such explanatory relations. The foregoing is an approximate account of the historically standard conception of coherence. While some proponents of coherentism have employed

Coherence theory of knowledge and justification

7

essentially this conception, others have in effect devised more idiosyncratic conceptions of coherence, conceptions whose connection to the historical concept is often quite tenuous. In particular, Rescher (1973, 1977) in fact employs both the standard conception of coherence, for certain purposes, and also a quite different concept that involves forming maximally consistent subsets of initially conflicting ‘data’ or ‘truth-candidates’, and then choosing among these subsets in a variety of ways that involve no appeal to standard coherence. And Lehrer (1974, 1990) has offered two subtly different versions of a general view that defines coherence in relation to the believer’s own subjective conception of probability or relative likelihood of truth; for a belief to cohere with the person’s system of beliefs is roughly for it to be judged to be more probable or more reasonable than any relevant competitor. The precise nature of coherence remains a largely unsolved problem. It is important to see, however, that difficulties in this area cannot yield anything like a decisive argument against coherence theories and in favour of their foundationalist rivals. This is so because the concept of coherence, or something so similar to it as to be capable of playing essentially the same role, is also an indispensable ingredient in virtually all foundationalist theories: coherence must seemingly be invoked to account for the relation between the basic or foundational beliefs and other non-foundational or ‘superstructure’ beliefs, in virtue of which the latter are justified in relation to the former. For this reason, giving an adequate account of coherence should not be regarded as exclusively or even primarily the responsibility of coherentists, despite the central role that the concept plays in their position.

4. Coherence and observation5 As mentioned above, few if any coherentists have wished to deny the seemingly obvious fact that sensory observation or perception plays a crucial role in justification (although they have not always been fully explicit on this point). It is thus incumbent on a coherence theory to explain how such observation can be construed in a non-foundationalist way. The central idea is that a belief that is produced by the senses, rather than being arrived at inferentially, might still depend for its justification on coherence with the background system of beliefs. But it is crucial here that the justification in question should still depend also in some way on the fact that the belief was a result of perception, since justification that depended only on the coherence of the belief’s propositional content with the rest of the cognitive system would make the observational status of the belief irrelevant. 5

BONJOUR, LAURENCE (1998). Knowledge and justification, coherence theory of. In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. Retrieved May 30, 2006, from http://www.rep.routledge.com/article/P009SECT4.

8

LAURENCE BONJOUR

One way to develop this idea is to focus on the fact that observational or perceptual beliefs are cognitively spontaneous; they simply strike the observer in an involuntary, coercive, non-inferential way, rather than as a product of any sort of inference or other discursive process, whether explicit or implicit. That a belief has this status, however, says nothing so far, according to the coherentist, about whether or how it is justified. Indeed, there is no reason to think that all cognitively spontaneous beliefs are justified, or even necessarily that most of them are, since the category would include hunches and irrational spontaneous convictions, as well as beliefs resulting from perception. But suppose that, as seems to be the case with most ordinary systems of belief, the system includes a belief to the effect that specific kinds of cognitively spontaneous beliefs (identified by their general subject matter, by their apparent mode of sensory production as reflected in the content of the belief, and by concomitant factors of various kinds) are, under specified (or perhaps ‘normal’) conditions, highly likely to be true. It then becomes possible to give a justifying reason for such a belief that appeals to its status as cognitively spontaneous and putatively observational, but still does so in a way that depends on the coherence with the background system of beliefs of the claim that a belief of this kind and produced in this way is true. Such a belief would be arrived at non-inferentially, but still justified by appeal to inference relations and coherence. (This view of observation is most explicit in BonJour (1985), but something like it seems implicit in Blanshard’s (1939) talk of ‘beliefs about the technique of acquiring beliefs’, in Sellars’s (1963) talk of ‘language-entry transitions’, in Quine’s talk of the ‘observational periphery’ of the ‘web of belief’, in Rescher’s (1973, 1974, 1977) idea of ‘data’, and in Lehrer’s (1974, 1990) discussion of a person’s trustworthiness in acquiring certain kinds of information). The foregoing provides at best only the beginning of a coherentist account of observation, leaving various problems to be solved that can only be touched on here. First, the other beliefs needed to give a justifying reason for a particular observational belief must themselves be justified in some fashion, without relapsing at this point into foundationalism. These beliefs will include at least (1) beliefs about the conditions; (2) the general belief about the reliability of the kind of cognitively spontaneous belief in question; and (3) beliefs about the occurrence of that particular belief, including the belief that it was indeed cognitively spontaneous. The justification for (1) will presumably have to include other observational beliefs, themselves justified in the same general fashion, so that any case of justified observation will normally or perhaps always involve a set of mutually supporting observations. The justification for (2) will appeal inductively to other cases of correct observation, as judged from within the system, as well as to more theoretical reasons for thinking that beliefs of the kind in question are generally produced in a reliable way. The justification for (3) will appeal to introspective beliefs, themselves constituting a species of observation, and ultimately to

Coherence theory of knowledge and justification

9

the believer’s comprehensive grasp of their overall system of beliefs – a grasp whose status poses one of the main problems to be considered below. Second, it is not enough for the justification of an observational belief that a reason of the foregoing sort should merely be present in a person’s system of beliefs, since such an individual might completely fail to notice that this was so, and might hold the belief on some other basis or for no reason at all. Thus even though the observational belief is not arrived at by inference, the availability of the inferential justification in question, even if never explicitly rehearsed, must be the reason why the believer continues to accept the belief and to appeal to it for further purposes. A full account of coherentist observation would have to spell out exactly what this requirement amounts to and how it can be satisfied. Third, the bare possibility of coherentist observation seems insufficient to accommodate the role that observation plays in our cognitive lives. Given the convictions that observation is not only possible but pervasive and that an appeal to observational evidence, whether direct or indirect, is essential for the justification of at least contingent beliefs about the world, an intuitively adequate coherence theory must somehow require and not just allow that a substantial observational element should be present in any justified system that includes such contingent beliefs. A view that insisted on such a requirement would thereby depart from a pure coherence theory, but might still avoid foundationalism if the coherentist account of observation is otherwise successful. (Such a requirement is relevant to several of the objections examined below.)

5. The standard objections6 In considering objections to coherence theories, we may begin with the three that are historically most standard and familiar. The first of these is what is commonly referred to as ‘the isolation problem’ or ‘the input objection’: an account of justification that appeals entirely to coherence within a system of beliefs seems to have the consequence that the justificatory status of the beliefs in the system will not depend in any way on the relation of the system to the world that it purports to describe, or on any sort of information derived from that world. This would seem to mean in turn that the truth of the component beliefs, if they happened to be true, could only be an accident, and thus that there is no reason to think that they are true and so no epistemic justification. The coherentist account of observation sketched above, if it can be successfully fleshed out, provides the beginning of an answer to this objection by showing how observational beliefs that are 6

BONJOUR, LAURENCE (1998). Knowledge and justification, coherence theory of. In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. Retrieved May 30, 2006, from http://www.rep.routledge.com/article/P009SECT5.

10

LAURENCE BONJOUR

apparently generated by the world might none the less be given a coherentist justification. In this way, a coherent system that involves a putatively observational component will at least seem from the inside to have input from the world and thus not to be isolated. Whether this seeming is likely to be veridical will depend, however, on the more general issue, discussed below, of whether and why coherentist justification should be regarded as conducive to finding the truth. The second familiar objection, already briefly alluded to earlier, is what may be called the alternative coherent systems objection: even given a relatively strong account of coherence, there will still be indefinitely many different possible systems of beliefs, each of which is as internally coherent as the others, and so all of which will be equally justified on a coherentist view – surely an absurd result. The response to this objection also depends crucially on the idea of observation. If, as suggested earlier, it will be a requirement for justification in an adequate coherence theory that there be a substantial observational component (that is, a substantial proportion of cognitively spontaneous beliefs that the system itself certifies as likely to be true and hence worthy of being accepted), then such alternative systems can no longer be freely invented and it is no longer obvious why they should be thought to exist. Only a system that is actually accepted and employed in cognitive practice can contain cognitively spontaneous beliefs and thus satisfy the requirement for observation. There is no way to guarantee that the acceptance of such beliefs will not lead quickly to incoherence in an arbitrarily devised system, even if it is initially coherent. (As this suggests, it is coherence over a period of time and not just at a moment that is ultimately the basis for justification in all coherence theories that have been seriously advocated.) The third of the standard objections is in effect a challenge to the coherentist to give a reason for thinking that adopting beliefs on the basis of coherentist justification is likely to lead to believing the truth. Different coherentists give very different responses to this crucial question, each problematic in its own way. These can only be briefly sketched here: (1) The absolute idealists in effect solve the problem by adopting a coherence theory of truth as well, thus reducing the gap between coherentist justification and truth (though only Blanshard is very explicit about this strategy). On such a view, truth is essentially identified with long-run justification, making it relatively easy to argue that seeking justified beliefs is likely to lead eventually to finding true ones – but at the significant cost of adopting an extremely implausible conception of truth. (2) Rescher (1973, 1974, 1977) attempts to give a pragmatic argument to the effect that the practical success which results from the employment of the coherent system makes it likely that the beliefs of the system are at least approximately true (in the sense of corresponding to independent reality). Unfortunately, however, the need for justification for the claims of practical success, which must presu-

Coherence theory of knowledge and justification

11

mably also be coherentist in character, threatens the project with vicious circularity. (3) BonJour (1985) attempts to give an a priori ‘metajustificatory argument’, relying on a rationalist and foundationalist conception of a priori knowledge, for the conclusion that a system of beliefs that remains coherent over the relatively long run, while receiving apparent observational input, is likely to be approximately true (again in the correspondence sense of truth). The main reason offered is that only approximate truth could explain continued coherence in the face of new observations. In addition to defending the general account of a priori justification presupposed, such an approach must also claim that sceptical explanations of one’s beliefs (for example their being produced by a Cartesian demon), are a priori less likely than the preferred explanation of correspondence with reality, a claim that many have found highly implausible. (4) Lehrer’s approach is to construct alternative conceptions of justification that involve the hypothetical replacement of erroneous beliefs in a person’s system of beliefs with their corrected alternatives, and then require that the person’s initially justified beliefs remain justified after such replacements in order for such beliefs to constitute knowledge. The main difficulty here is that such an approach seems to concede that ‘personal justification’ – the sort of justification which exists before the hypothetical replacements – is not in itself conducive to finding the truth, even though such personal justification is the only sort that the believer is in general ever actually aware of.

6. The problem of access7 In addition to the foregoing objections, there are a number of further problems with which an adequate coherence theory would have to deal. Perhaps the most urgent of these is that of whether coherentist justification is accessible to the believer in the way that it must be if an internalist position is to result. Assuming for the moment, as is the case with all the positions discussed here, that coherentist justification is taken to require coherence with the believer’s entire system of beliefs, then there are three aspects to this problem: (1) whether the believer has adequate access to their system of beliefs; (2) whether the believer has an adequate grasp of the concept of coherence; and (3) whether the believer is able to apply the concept of coherence to their system of beliefs in a way that will yield a definite assessment. All these aspects pose serious problems, and (3) in particular is anything but trivial, even given satisfactory solutions to (1) and (2). But (1) is the most difficult and will accordingly be the main focus here. A believer’s access to their own system of beliefs is in fact seriously 7

BONJOUR, LAURENCE (1998). Knowledge and justification, coherence theory of. In E. Craig (Ed.), Routledge Encyclopedia of Philosophy. London: Routledge. Retrieved May 30, 2006, from http://www.rep.routledge.com/article/P009SECT6.

12

LAURENCE BONJOUR

problematic in two quite different ways. First, there is the problem of the epistemological status of the result of such access if it were achieved, which we may think of as a reflective meta-belief describing the entire contents of the system. Such a meta-belief would be clearly contingent and empirical and hence one that on any coherence theory of the sort under consideration here ought to be itself justified by appeal to coherence. But since any coherentist justification that is to be accessible to the believer must appeal to such a meta-belief in order to characterize the system of beliefs with which the belief to be justified must cohere, a coherentist justification of that meta-belief itself appears to be totally and irrevocably circular (and no appeal to non-linear justification will help here, since what is being explained is how the very sort of non-linear justification advocated by the coherentist is possible). The most explicit discussion of this issue is given by BonJour (1985), who responds by invoking what he calls ‘the Doxastic Presumption’. The idea is that coherentist justification must presume that the believer’s grasp of their overall system of beliefs is at least approximately correct (small corrections being possible by appeal to coherence). This has the consequence that the resulting justification is contingent upon the presumed correctness of this grasp, and hence that there is no possible answer on the part of a coherence theory to the specific variety of scepticism that questions whether this presumption is indeed correct. This has seemed to many to be a very drastic result, but it is unclear what the alternative might be, so long as foundationalism is eschewed. Even if the foregoing issue were resolved in a satisfactory way, there is still the second aspect of the present problem: the quite sticky issue of whether ordinary believers ever in fact possess or could possess anything like the reflective grasp of the entire contents of their systems of beliefs that a coherence theory seems to require. On this issue, it seems likely that a coherence theory will have to concede that ordinary cases of justification are at best only an approximation, and perhaps a fairly distant one, to the ideal justification that a coherence theory portrays. One further problem is worth a brief mention. If, as is almost always the case, a coherence theory appeals to coherence over the relatively long run, then the issue arises of how the memory beliefs upon which any access to the fact of continued coherence must seemingly rely, are themselves to be justified. Many philosophers have offered coherence theories of the justification of memory beliefs, but such a view is again threatened with vicious circularity if the only reason for thinking that coherentist justification is conductive to truth – and so that the memory beliefs in particular are true – relies on coherence over time, and so on those very memory beliefs themselves.