Testimony and Disagreement

You know your date of birth. You were present at the event and remember nothing of it. You know it because you were told, by people you had no means of checking, and the documents that confirm it were produced by other people you have never met.

Most of what anyone knows has this structure. The shape of the continents, the existence of atoms, the history of any country, the contents of the medicine cabinet: nearly all of it arrives by someone saying so. The beginner tier listed testimony among the sources of knowledge. This article asks the two questions that listing postponed. What entitles you to believe what you are told? And what should happen to your confidence when someone equally careful tells you the opposite?

The first question

Two positions divide the field, and they divide it along a line the internalism debate has already drawn.

Reductionism descends from Hume, who held that our confidence in testimony derives entirely from having observed that what people say tends to match what happens. Its appeal is that it refuses to make credulity a virtue. Believing whatever anyone tells you is a recipe for error, and it seems the hearer ought to have some reason for trust beyond the fact that words were spoken.

The difficulty is coverage. Global reductionism, which requires independent confirmation that testimony in general is reliable, cannot get started. C. A. J. Coady's 1992 book pressed this point decisively: the observations one would need are themselves overwhelmingly testimonial. You have not personally checked that the reports of distant events tend to be accurate, or that encyclopaedias match the world, and the evidence you would cite for their reliability reached you through further reports. Children face the problem in its purest form, since they acquire language and a picture of the world from testimony before they are in any position to evaluate speakers.

Local reductionism retreats to requiring reasons for trusting a particular speaker on a particular occasion, and this is more plausible. But the reasons it can supply are thin. You trust a stranger who tells you the time. What independent evidence of their reliability do you have? That people generally report the time accurately, which is a general claim about testimony, and so the local version borrows from the global one it abandoned.

Anti-reductionism, associated with Thomas Reid and developed in contemporary form by Tyler Burge, treats trust as a default. Reid held that we are constituted with a principle of credulity that matches a principle of veracity in speakers, and that the two together make communication possible. Burge's version is that a hearer is entitled to accept what is presented as true by a rational source, since the capacity for rational communication is itself a guide to truth.

The objection is gullibility. If the default is acceptance, the anti-reductionist must say that the credulous child and the careful adult are equally entitled to believe a skilled liar whose lie contains no internal warning signs. Jennifer Lackey's response is that both theories go wrong by locating testimonial knowledge entirely in the hearer. What matters is also whether the speaker's testimony is reliable, and she argues that testimony can transmit knowledge a speaker lacks, as when a teacher who does not believe in evolution nonetheless teaches it accurately.

The second question

The testimony dispute assumes a speaker who knows more than the hearer. The harder case is when they know the same.

The stock example is due to David Christensen. You and a friend split a restaurant bill, each calculating the share in your head. You get forty-three pounds. Your friend gets forty-five. You have dined together many times and have equally good track records at this. What should you now believe?

Almost everyone says you should lose confidence. You should not insist on forty-three, and you should not switch to forty-five. You should suspend judgement until someone uses a calculator. The question is what principle generates that verdict and how far it extends.

The argument works cleanly on the restaurant bill, and conciliationists such as Christensen and Adam Elga hold that it works everywhere. The consequence is striking. On most contested philosophical, political, and religious questions, there are people who are, by any honest measure, your peers and who disagree. Conciliationism therefore recommends widespread suspension of judgement on exactly the questions people care most about.

Thomas Kelly's defence of steadfastness turns on a distinction between the two parties. Of two peers who disagree, one has in fact assessed the evidence correctly. The evidence supports that person's view. Why should the person who got it right be required to abandon a correctly reached conclusion merely because someone else got it wrong? The symmetry the conciliationist appeals to is symmetry from the outside. From the inside, one of the parties has the evidence on their side, and the evidence does not stop supporting a conclusion because a peer misjudged it.

The standard reply is that the correct party cannot tell that they are the correct party. They have exactly the same grounds for confidence as their mistaken peer. Kelly's total evidence view then tries to split the difference: the first-order evidence and the higher-order evidence of disagreement both count, and how much each counts depends on how strong the first-order case was. When the evidence is overwhelming, as when a peer insists that two plus two is five, you may stay put and conclude your peer has blundered.

How the two questions connect

Testimony and disagreement look like separate topics, and they are the same topic at different points on a scale. When someone who knows much more than you tells you something, you defer. When someone who knows the same tells you the opposite, you face disagreement. When someone who knows less tells you something false, you may reasonably ignore them. What varies is your estimate of the other person's reliability relative to your own, and the norms governing each case are points on a single continuum of how much weight another mind's output should carry.

That makes both questions questions about the social dimension of knowledge, which the individualist tradition from Descartes onward largely ignored. The final article of this tier makes that dimension its subject.

Next

What goes wrong when the weight given to a speaker's word is distorted by prejudice about who they are, and why that wrong is specifically epistemic.