coherentism in epistemology, a theory of the structure of knowledge or justified beliefs according to which all beliefs representing knowledge are known or justified in virtue of their relations to other beliefs, specifically, in virtue of belonging to a coherent system of beliefs. Assuming that the orthodox account of knowledge is correct at least in maintaining that justified true belief is necessary for knowledge, we can identify two kinds of coherence theories of knowledge: those that are coherentist merely in virtue of incorporating a coherence theory of justification, and those that are doubly coherentist because they account for both justification and truth in terms of coherence. What follows will focus on coherence theories of justification.
Historically, coherentism is the most significant alternative to foundationalism. The latter holds that some beliefs, basic or foundational beliefs, are justified apart from their relations to other beliefs, while all other beliefs derive their justification from that of foundational beliefs. Foundationalism portrays justification as having a structure like that of a building, with certain beliefs serving as the foundations and all other beliefs supported by them. Coherentism rejects this image and pictures justification as having the structure of a raft. Justified beliefs, like the planks that make up a raft, mutually support one another. This picture of the coherence theory is due to the positivist Otto Neurath. Among the positivists, Hempel shared Neurath’s sympathy for coherentism. Other defenders of coherentism from the late nineteenth and early twentieth centuries were idealists, e.g., Bradley, Bosanquet, and Brand Blanshard. (Idealists often held the sort of double coherence theory mentioned above.)
The contrast between foundationalism and coherentism is commonly developed in terms of the regress argument. If we are asked what justifies one of our beliefs, we characteristically answer by citing some other belief that supports it, e.g., logically or probabilistically. If we are asked about this second belief, we are likely to cite a third belief, and so on. There are three shapes such an evidential chain might have: it could go on forever, if could eventually end in some belief, or it could loop back upon itself, i.e., eventually contain again a belief that had occurred ‘higher up’ on the chain. Assuming that infinite chains are not really possible, we are left with a choice between chains that end and circular chains. According to foundationalists, evidential chains must eventually end with a foundational belief that is justified, if the belief at the beginning of the chain is to be justified. Coherentists are then portrayed as holding that circular chains can yield justified beliefs. This portrayal is, in a way, correct. But it is also misleading since it suggests that the disagreement between coherentism and foundationalism is best understood as concerning only the structure of evidential chains. Talk of evidential chains in which beliefs that are further down on the chain are responsible for beliefs that are higher up naturally suggests the idea that just as real chains transfer forces, evidential chains transfer justification. Foundationalism then sounds like a real possibility. Foundational beliefs already have justification, and evidential chains serve to pass the justification along to other beliefs. But coherentism seems to be a nonstarter, for if no belief in the chain is justified to begin with, there is nothing to pass along. Altering the metaphor, we might say that coherentism seems about as likely to succeed as a bucket brigade that does not end at a well, but simply moves around in a circle. The coherentist seeks to dispel this appearance by pointing out that the primary function of evidential chains is not to transfer epistemic status, such as justification, from belief to belief. Indeed, beliefs are not the primary locus of justification. Rather, it is whole systems of belief that are justified or not in the primary sense; individual beliefs are justified in virtue of their membership in an appropriately structured system of beliefs. Accordingly, what the coherentist claims is that the appropriate sorts of evidential chains, which will be circular – indeed, will likely contain numerous circles – constitute justified systems of belief. The individual beliefs within such a system are themselves justified in virtue of their place in the entire system and not because this status is passed on to them from beliefs further down some evidential chain in which they figure. One can, therefore, view coherentism with considerable accuracy as a version of foundationalism that holds all beliefs to be foundational. From this perspective, the difference between coherentism and traditional foundationalism has to do with what accounts for the epistemic status of foundational beliefs, with traditional foundationalism holding that such beliefs can be justified in various ways, e.g., by perception or reason, while coherentism insists that the only way such beliefs can be justified is by being a member of an appropriately structured system of beliefs.
One outstanding problem the coherentist faces is to specify exactly what constitutes a coherent system of beliefs. Coherence clearly must involve much more than mere absence of mutually contradictory beliefs. One way in which beliefs can be logically consistent is by concerning completely unrelated matters, but such a consistent system of beliefs would not embody the sort of mutual support that constitutes the core idea of coherentism. Moreover, one might question whether logical consistency is even necessary for coherence, e.g., on the basis of the preface paradox. Similar points can be made regarding efforts to begin an account of coherence with the idea that beliefs and degrees of belief must correspond to the probability calculus. So although it is difficult to avoid thinking that such formal features as logical and probabilistic consistency are significantly involved in coherence, it is not clear exactly how they are involved. An account of coherence can be drawn more directly from the following intuitive idea: a coherent system of belief is one in which each belief is epistemically supported by the others, where various types of epistemic support are recognized, e.g., deductive or inductive arguments, or inferences to the best explanation. There are, however, at least two problems this suggestion does not address. First, since very small sets of beliefs can be mutually supporting, the coherentist needs to say something about the scope a system of beliefs must have to exhibit the sort of coherence required for justification. Second, given the possibility of small sets of mutually supportive beliefs, it is apparently possible to build a system of very broad scope out of such small sets of mutually supportive beliefs by mere conjunction, i.e., without forging any significant support relations among them. Yet, since the interrelatedness of all truths does not seem discoverable by analyzing the concept of justification, the coherentist cannot rule out epistemically isolated subsystems of belief entirely. So the coherentist must say what sorts of isolated subsystems of belief are compatible with coherence.
The difficulties involved in specifying a more precise concept of coherence should not be pressed too vigorously against the coherentist. For one thing, most foundationalists have been forced to grant coherence a significant role within their accounts of justification, so no dialectical advantage can be gained by pressing them. Moreover, only a little reflection is needed to see that nearly all the difficulties involved in specifying coherence are manifestations within a specific context of quite general philosophical problems concerning such matters as induction, explanation, theory choice, the nature of epistemic support, etc. They are, then, problems that are faced by logicians, philosophers of science, and epistemologists quite generally, regardless of whether they are sympathetic to coherentism. Coherentism faces a number of serious objections. Since according to coherentism justification is determined solely by the relations among beliefs, it does not seem to be capable of taking us outside the circle of our beliefs. This fact gives rise to complaints that coherentism cannot allow for any input from external reality, e.g., via perception, and that it can neither guarantee nor even claim that it is likely that coherent systems of belief will make contact with such reality or contain true beliefs. And while it is widely granted that justified false beliefs are possible, it is just as widely accepted that there is an important connection between justification and truth, a connection that rules out accounts according to which justification is not truth-conducive. These abstractly formulated complaints can be made more vivid, in the case of the former, by imagining a person with a coherent system of beliefs that becomes frozen, and fails to change in the face of ongoing sensory experience; and in the case of the latter, by pointing out that, barring an unexpected account of coherence, it seems that a wide variety of coherent systems of belief are possible, systems that are largely disjoint or even incompatible. See also COHERENCE THEORY OF TRUTH, EPISTEMOLOGY, FOUNDATIONALISM , JUSTIFI – CATIO. M.R.D.