Knowledge from testimony
A Socratic walk-through of knowledge from testimony — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why is it reasonable to believe almost everything you know without ever checking it yourself?
Take an inventory of what you know and sort it by how you came to know it. A thin pile comes from your own eyes and hands: the taste of this coffee, the layout of your street. Everything else — your date of birth, the existence of Peru, the structure of DNA, what happened in 1789, that the medicine on your shelf does what the label says — arrived through somebody telling you, or writing it down. You have checked essentially none of it.
That ought to worry us more than it does. The standard picture of a justified belief has the believer holding some evidence. Here the evidence is that someone said so, which looks like the weakest thing in the room. So what entitles you to almost your entire body of knowledge?
Reasoning it through
REASONING #The obvious first move is to say testimony has earned its place. You have, over a lifetime, seen countless claims from other people turn out to be true — weather forecasts, timetables, directions to the station — so induction licenses a general confidence in what people say, adjusted by source. Call that the reductionist answer: testimony is not a basic source of knowledge at all, but a derived one, reducible to your own perception, memory and inference. Hume argued in essentially this form, treating the credit we give to witnesses as founded on our experience of how far testimony has conformed to fact.
It is a tidy answer. Try, though, to actually run the audit it describes. To establish that testimony has a good track record, you must know, in enough cases, both what was said and what was true. How did you learn what was true? For nearly everything, by more testimony. The map that confirmed the directions, the almanac that confirmed the forecast, the calibration of the instrument you trusted — each is a report by someone else. C. A. J. Coady pressed exactly this point: the reductionist programme cannot get off the ground, because the only evidence base large enough to support the induction is already saturated with the thing it was meant to justify.
Push further and it gets worse. Your own perception is only usable because you were taught what you are looking at; your memory is dated by a calendar you did not verify; the very language in which the audit is conducted was learned by taking others' word for what words mean. There is no vantage point outside the testimony system from which to grade it.
So consider the other option. Thomas Reid proposed a pair of native dispositions — a veracity principle inclining us to speak truly, and a credulity principle inclining us to believe what we are told — and treated testimony as a basic source of knowledge alongside perception and memory, needing no external certificate. On that account you have a default entitlement to believe what you are told. Not a guarantee: an entitlement that any specific reason for doubt cancels.
Does that make us gullible? Watch what people actually do and the answer is no. We do not check claims, but we do run continuous cheap monitoring: does this fit what I already hold, does the speaker have an obvious interest, has this source been wrong before, do independent sources agree. Dan Sperber and colleagues called this epistemic vigilance, and it starts early — preschool children who have seen an adult name familiar objects wrongly will subsequently prefer a different adult's word for a new one. That is not verification. It is filtering.
Notice, then, what the whole arrangement buys. If knowledge could not be transmitted on trust, each person's holdings would be capped at what one lifetime of first-hand inquiry could establish, and no cumulative body of knowledge could exist at all. Trust is not a regrettable shortcut around the ideal of checking everything. It is the condition of there being much to know.
The analogy
THE ANALOGY #Testimony works like the currency in your pocket. You accept the note without assaying it, and you can do that only because a whole system — issuers, courts, other people's willingness to take it from you in turn — stands behind it. Your confidence is not really in the note; it is in the network, which you also have not inspected.
a note's value is created by everyone accepting it, whereas a claim's truth is not — unanimous acceptance of a falsehood leaves it false, so a testimonial system can be widely trusted, internally consistent, and wrong.
Clarifying the model
THE MODEL #The two positions are less far apart in practice than in principle. Elizabeth Fricker's local reductionism concedes that we cannot justify testimony wholesale but insists the hearer must have some grounds about this speaker; anti-reductionists reply that even those grounds are usually testimonial. Jennifer Lackey's dualism argues both sides are needed — the speaker must actually be reliable, and the hearer must have positive reason to take her as such.
What survives all of it is the negative result, and it is the honest core here: no purely reductionist account can succeed, because the audit it demands is circular. Whatever the correct positive theory, dependence on others is not a defect in our knowledge that better epistemic hygiene would remove.
Two corrections worth making. First, "you cannot check it yourself" does not mean nothing checks it — science, journalism and law are institutions built to make someone's checking bear on what the rest of us are told, and their health is what your entitlement actually rests on. Second, entitlement is defeasible by design, which is why the interesting failures are not credulity but corrupted channels: a source you have good reason to trust, reporting the interests of somebody you have never heard of.
A picture of it
THE PICTURE #How to readStart at the slanted node, where a claim arrives. It meets vigilance first — cheap monitoring that can withhold belief, but does not verify anything. Almost all traffic then takes the left branch out of the diamond, arriving at belief that any later reason for doubt can cancel. The right branch is the reductionist's route: go and check. Follow it and it lands on the loop back to the start, which is the argument's whole point — every instrument of checking is another person's report, so the path never leaves the system it was trying to grade from outside.
What became clearer
WHAT CLEARED #Almost everything you know rests on trust, and this is not a lapse to be corrected: the alternative programme, verifying your way to a track record for testimony, cannot be carried out, because its verification runs on testimony. What we actually do is accept by default and monitor cheaply — and what makes that reasonable is less the individual's judgement than the institutions that make somebody else's checking count.
Where to go next
ONWARD #- How expertise is identified by non-experts, when assessing the expert's reasoning is what you lack.
- What goes wrong when independent sources are only apparently independent, and agreement stops being evidence.
Key terms
TERMS #| Term | What it means |
|---|---|
| Testimony | a claim believed on the ground that someone asserted it, in speech or writing. |
| Reductionism (about testimony) | the view that testimony's justification derives from the hearer's own perception, memory and inductive track record. |
| Anti-reductionism | the view that testimony is a basic source of justification, giving a default entitlement that reasons for doubt defeat. |
| Defeater | a specific reason to doubt a claim or its source, which cancels the default entitlement to believe it. |
| Epistemic vigilance | the low-cost filtering people apply to what they are told, short of independent verification. |
Every term the collection defines is gathered in the glossary.