Sources: Perception, Memory, Testimony, Reason

Pick any belief you hold and ask where it came from. The answer will be one of a small number of things: you saw it, you remember it, someone told you, or you worked it out. Almost nothing falls outside that list, which is convenient, because it means the general question of how beliefs get justified can be broken into four smaller questions about four sources, each of which fails in its own characteristic way.

Perception

Perception is the source that feels least like a source. When you see the cup on the desk, no reasoning is experienced, and the belief arrives already formed. That immediacy is why perception is the model for every account of basic, non-inferential justification.

It fails in three ways, and they are worth keeping separate.

The first is illusion: the stick in water looks bent, the two lines of equal length look unequal, the distant tower looks round when it is square. The senses deliver something the world does not match, under conditions that are perfectly ordinary.

The second is hallucination, where there is no object at all and the experience is indistinguishable from the real thing to the person having it. The possibility that an experience could be exactly as it is with nothing before you is what makes the philosophy of mind's argument from illusion pressing, and that argument is taken up in that subject rather than here.

The third is more insidious because it does not feel like error. What a person sees is shaped by what they expect: the radiologist sees a tumour on the film where the layman sees grey, and the witness recalls a weapon that was never there. Perception is not a neutral recording device, and the fact that it is trained is exactly what makes the expert's perception valuable and the biased observer's unreliable.

Memory

Nearly everything you believe right now is believed on this basis. You are not currently perceiving that Ottawa is the capital of Canada, and you are not reasoning it out. You remember it.

A question about memory's role divides the literature and is easy to state. Does memory merely preserve a justification acquired earlier, so that the belief is as well supported now as it was when it was formed and no better? Or can memory generate justification of its own, so that a vivid seeming recollection counts in favour of what it presents even when the original grounds are long gone?

The preserving view has the attraction of tidiness and an awkward consequence. Almost nobody can produce the original grounds for most of what they remember. You believe the capital fact and you have no idea who told you. On the strict preserving view your justification is whatever that forgotten source supplied, which you cannot now assess.

The generating view fits the practice better and has to reckon with what psychology has established about how memory works. Remembering is reconstructive rather than retrieval from storage, and reconstruction can insert material that was never there. Confident, detailed, emotionally vivid recollections are sometimes wholly false, and the confidence does not track the accuracy. A source that generates justification in proportion to vividness would generate it in exactly the cases that go wrong.

Testimony

The scale of dependence here is easy to underestimate. Your name, your date of birth, the existence of countries you have not visited, the germ theory of disease: all testimony, none of it checked. A person restricted to what they had perceived and worked out for themselves would know almost nothing.

The philosophical question is what entitles the acceptance. Two answers have been given, and they descend from a genuine disagreement between two eighteenth-century Scots.

David Hume treated testimony as a special case of ordinary inference from experience. You have observed that reports of certain kinds tend to be true, and that observation justifies accepting further reports of those kinds. On this account testimony is not a basic source at all: it reduces to perception, memory, and inference.

Thomas Reid rejected the reduction. He held that human beings have two innate dispositions, a propensity to speak the truth and a disposition to trust what they are told, and that the trust comes first rather than being earned. A child accepts what it is told long before it has assembled any track record, and if it did not, it could never acquire the language in which the track record would be stated.

Reason

Reason covers two rather different jobs. One is inference, taking you from things you already believe to further things that follow. The other is the direct grasp of certain truths without inference, such as that nothing is both round and not round, or that if one thing is larger than a second and the second larger than a third then the first is larger than the third.

The second job is the contested one. It has seemed to many philosophers that some things can be seen to be true by thinking alone, and this apparent capacity does most of the work in mathematics and logic. The intermediate tier examines it under the heading of a priori justification and asks what such grasping could consist in.

Its failure mode is that things can seem self-evident and not be. For two thousand years Euclid's parallel postulate looked like an obvious truth about space that ought to be derivable from the others. It is neither derivable from them nor forced on us, and consistent geometries exist in which it fails. Nothing about how obvious a claim feels guarantees that it holds.

A fifth source belongs on the list and is treated elsewhere. Introspection, the awareness a person has of their own current mental states, has often been thought the most secure source of all, and the philosophy of mind ladder examines both the case for that and the experimental results that undercut it.