Coherentism is a theory of justification that claims that judgements are justified by their coherence with other beliefs we already hold. It does not oblige us to enquire how those previous beliefs were in turn justified (which would lead to the infinite regress problem), but only to consider the coherence of new beliefs in the present time. So, if our sense were to throw up a completely incoherent experience (say, a bear writing a philosophy book), or if a priori reasoning were to lead me to a conclusion that contradicts other such reasoning (say, that 2+2=5), then I could justifiably reject these claims.

Coherentism has adopted a range of different accounts of what exactly is meant by ‘coherent’. At root it must mean logically non-contradictory, but it can also incorporate requirements for probabilistic or explanatory consistency. A bear writing a philosophy book is not contradictory, but it would be highly improbable within my existing framework of beliefs, because previous experience suggests that philosophy books are only written by humans and that bears show no signs of literacy. The things I consider probable within my framework would, to be consistent, make a bear writing a philosophy book extremely improbable. The hypothesis of a bear writing it would also not be explanatorily consistent, as it would not explain the features of the philosophy book in ways that the hypothesis of a human doing so would. Coherentism would generally require, then, that my beliefs be free not only of contradiction but also of extreme improbability and unfruitful explanation.

Coherentism incorporates all new evidence from the senses so long as that evidence is coherent. As a theory of justification, it is concerned with how we adopt experience into belief rather than with the belief itself. Clearly some aspects of belief will be rejected as incoherent following coherentism (e.g. hallucinations, mistaken calculations, conspiracy theories), whilst most will be accepted (e.g. everyday experiences within the realm of reasonable probability, for which feasible explanations exist). I might accept that there is a tiger running down an English street (coherent explanation – it has escaped from a wildlife park) but not a dragon (too improbable on the basis of known animals).

Nevertheless, coherentism does not seem to stand up to scepticism any better than positive foundationalism does. A set of entirely logically consistent, consistently probable, and mutually explanatory beliefs could perfectly well all be wrong – an extended construction of deluded fantasy. Anton’s Syndrome, where a person who has become blind denies their blindness and describes a perfectly coherent (but from our point of view, false) visual world all around them, is a complete example of this[1]. For this reason we cannot accept coherentism by itself as a sufficient basis for justification.

However, we do also need our beliefs to be coherent. For all that I don’t take mathematical claims to be absolute, they are generally correct within the sphere we experience. We would not be justified in believing that 2+2=5 even if God apparently said so. There may also be places where bears write philosophy books, bus conductors suddenly turn into giant parrots, and the leader of the Conservative Party is an aardvark, but not in the world I think I inhabit. Coherence seems to be necessary for our beliefs to be justified, even though it is not sufficient.

Another difficulty with coherentism is its apparent dependence on convention. Beliefs that are taken to be coherent are those accepted by most of the people around us, because the vast majority of what we take to be our knowledge is indirectly justified by the experience of others. Our ideas of what is probable or of what is explanatory depend largely on reported experience. Even our ideas about what is logically consistent often depend on what others take to be logically consistent, or on what constitutes acceptable premises for reasoning, rather than what we have worked out for ourselves. For example, the tradition of the negative implications of Pyrrhonian scepticism has been conventionally accepted since the Renaissance rediscovery of scepticism, even though there is no contradiction between Pyrrhonism and provisional belief. Even where the reasoning is unquestionable, any reasoning is only as good as its premises, and those of coherentism will unavoidably be conventional.

In schools of thought that rely on coherentism, the absence of a foundational justification is taken to be a reason to rely on convention instead – as in the relativist approach to ethics, where the lack of a universal ethics leaves us with a localised social or cultural ethics. However, the application of analysis to conventional beliefs can still reveal them to be contradictory. For example, it is quite common to make claims like “It’s just my opinion, but abortion is absolutely wrong”, or “Everyone has their own personal God”, which attempt to combine absolute assumptions with conventional ones. Convention can be incoherent just because it does not incorporate awareness of its own limits.

So, we do not need to reject coherentism as necessary on the grounds that it can be used to support conventionalist assumptions, but rather keep up an awareness of the conventionalist assumptions that might be smuggled into coherentist reasoning. Those conventionalist assumptions will undermine coherentism because they do not incorporate awareness of their own limits, but coherentism could still work without them by considering all issues of coherence critically rather than merely in acceptance of convention.

If we accept coherence as necessary, we do not have to accept coherentism as a complete explanation of justification, because we do not accept that coherence is sufficient for justification. To identify the other necessary conditions for justification, then, we must go back and reconsider foundationalism.

[1] Schulz (2010) pp. 67-8