Foundationalism, Coherentism, Infinitism

The regress argument closed off the field before anyone had a chance to survey it. Once the chain of reasons must either stop, loop, or run on without end, the theories of justification are not an open list. There are three, plus scepticism, and the interest lies in what each has to pay.

This article takes them in that order. The point of treating them together is that none of them is best understood on its own terms. Each is a bet that one horn of the trilemma is survivable and that the other two are not, and the arguments for any one of them are largely arguments against the others.

What the theories are theories of

Before the three positions, one distinction that all of them need.

The distinction matters because a belief can be supported by excellent reasons that play no part in why the person has it. Someone who believes a conspiracy theory for bad reasons does not acquire justification when a good argument for the same conclusion turns up in a journal they have never read. The structural theories are theories about what chains of basing look like, not about what arguments happen to exist.

This is also where the beginner tier's distinction between propositional and doxastic justification is doing work. A proposition can be justified for you while your belief in it is not, and the difference is whether the basing relation runs through the right things.

Foundationalism

The position is the oldest and still the most widely held. Its structure is an inverted pyramid: a base that supports itself, and everything else resting on the base through relations of support that run in one direction only.

The classical version, running from Descartes through the logical empiricists, made the base very small and very secure. Basic beliefs were about one's own current mental states, or were simple self-evident truths, and they were taken to be incorrigible. The appeal is obvious. If the foundations cannot be wrong, the third horn is refused outright.

The cost is equally obvious, and it sank the classical version. A base of that kind is too thin to support anything. From the incorrigible fact that you are having a reddish visual experience, nothing about the external world follows deductively, and the inductive bridges that were proposed to get there were never convincing. The classical foundationalist ends up with a secure base and no building.

Modest foundationalism, which is what most contemporary foundationalists hold, gives up incorrigibility. Basic beliefs are defeasibly justified: perceptual beliefs about the immediate environment count as basic, they can be overturned by further evidence, and they confer justification that is real but less than certain. This buys a base wide enough to build on and gives up the hope of refuting the sceptic outright.

The standing problem for either version is stated most sharply by Wilfrid Sellars, and the intermediate reader should meet it here even though its full development belongs elsewhere. A basic belief is supposed to be justified by experience rather than by another belief. But experience justifies a belief only if experience has content that can stand in logical relations to it, and anything with that kind of content looks like something that can itself be assessed as correct or incorrect, which is to say it looks like a belief after all. Sellars calls the assumption that there is a layer of experience both contentful enough to justify and innocent enough to need no justification the Myth of the Given.

Coherentism

The coherentist rejects the picture rather than a premise within it. There is no chain, so the question of where it terminates does not arise. Beliefs are justified in bulk, by membership in a set that hangs together, and support runs in every direction at once.

The linear circularity that makes the second horn look hopeless is not what the coherentist is defending. Nobody holds that you may justify A by B and B by A and call it a day. The claim is that justification is holistic, so that the unit of assessment is the whole set, and that the apparent circle is an artefact of trying to describe a holistic relation one link at a time.

The problem the position cannot shake is called the isolation objection, and the sample version of it has been called the alternative systems objection. Coherence is a relation among beliefs. A well-constructed fiction is coherent. A paranoid delusion of sufficient sophistication is coherent, and is often more coherent than the belief set of someone responsive to messy evidence. If coherence is all there is to justification, then nothing about a coherent system connects it to the world, and there will be indefinitely many equally coherent systems that contradict one another.

The standard coherentist answer is to insist that a belief set must include beliefs about the reliability of its own sources, and that a system which coheres while containing no such beliefs is not maximally coherent. Laurence BonJour developed a version of this in 1985 and then, in a rare and instructive move, abandoned coherentism about a decade later on the grounds that the answer does not work: a system can contain the belief that its perceptual sources are reliable, and cohere beautifully, without those sources being reliable.

Infinitism

Infinitism has one serious contemporary defender, Peter Klein, and is worth taking seriously for reasons beyond completeness.

The objection everyone raises first is that finite minds do not contain infinitely many beliefs. Klein grants it. His claim is not that the believer has traversed or even entertained the series, but that the reasons are available, in the sense that the person is capable of producing the next one when the demand for it is pressed. Availability rather than possession is the operative notion, and a finite mind can stand in that relation to an infinite series in the same way that a person who can add can be said to have infinitely many sums available.

The positive argument is that the other two horns fail and that infinitism is what remains. Foundationalism must say that some belief is justified though there is nothing further to be said for it, which is arbitrariness with a name. Coherentism must say that a belief can appear among its own grounds. Infinitism alone avoids both, and it does so by paying a price that is strange rather than incoherent.

Whether the second premise is fair is the crux. The modest foundationalist denies that stopping at a perceptual belief is arbitrary, since the stopping point is fixed by experience rather than by fiat. Klein's reply is that this either makes experience into a reason, in which case the chain continues and foundationalism collapses into his view, or leaves the stopping point unexplained after all.

Scoring the three

The honest summary is that no position has won and that the reasons are structural.

Foundationalism has the base rate. It matches the phenomenology of ordinary justification, where reasons do seem to run out, and it can explain how a finite creature gets justified beliefs. It owes an account of the base that neither collapses into the Myth of the Given nor admits so much that the term stops meaning anything.

Coherentism has the better account of revision. Belief systems really do get adjusted holistically, and the history of science looks far more like Neurath's boat than like construction on bedrock. It owes a connection to the world.

Infinitism has the cleanest response to the trilemma and the least plausible psychology. It owes an account of availability that is demanding enough to be justification and loose enough that a finite believer satisfies it.

What has moved the field is not a resolution but a change of question. Reliabilism, in the next article but one, asks what makes a belief-forming process good rather than what supports what, and in doing so declines to enter the structural dispute at all.

Next

The dispute the objection just anticipated: whether what justifies a belief has to be something the believer can access, or whether it can be a fact about the world the believer knows nothing about.