On Knowing

Ask most people how they know something and they will describe a chain of reasoning that led them there. The order feels obvious: the evidence comes in, the mind weighs it, and knowledge settles at the end like sediment.

Robert Burton, a neurologist, wrote a book, On Being Certain, arguing that the real order is closer to the reverse. The feeling that we know arrives first, unbidden, and the reasoning is assembled afterward to account for a conviction we already have.

He means this literally. The sense of knowing is a mental sensation, produced by machinery we don’t control. It’s in the same family as the feeling of a name on the tip of the tongue or the click of recognising a face in a crowd. It can fire correctly, and it can fire when we are plainly wrong, feeling exactly the same either way. That last point is the one that matters, because it means the strength of a conviction tells you nothing reliable about whether the conviction is true.

The clearest evidence for this comes from memory rather than from argument. After a public shock such as the Challenger explosion or the morning of September 11, psychologists have asked people to write down where they were and what they were doing, then returned years later and asked again. A striking share of the later accounts are wrong, sometimes wrong in every particular. What doesn’t fade is the certainty. People will describe a scene they have demonstrably invented and report complete confidence that they are remembering it correctly. The feeling of knowing and the fact of knowing have come apart, and the person can’t tell from the inside that they have.

If certainty is a sensation, the puzzle is why it ever tracks reality at all, but sometimes it does. Consider the areas of expertise where being wrong has a short, hard leash on it. An engineer’s certainty that a beam will hold is checked by whether the beam holds. A doctor’s judgement is checked by whether the patient recovers. A pilot’s is checked, unforgivingly, by the ground. Each works inside a loop where the world answers back quickly, and a wrong feeling of knowing is corrected before it can harden. Over years, that correction drags the sensation into rough alignment with what is actually so. The certainty of people working inside such a loop becomes trustworthy, not because they feel it more carefully than the rest of us, but because something outside them keeps grading it.

Now take the leash away. In history, in questions of authorship, in politics, in taste, a wrong certainty costs nothing and is corrected by nothing. The feeling floats free. It can grow as strong as any engineer’s certainty, often stronger precisely because nothing ever contradicts it, while keeping no contact with the facts at all. This is why the most confident voice in a room about the distant past is so often the least constrained by the evidence: confidence and constraint have come unhooked from each other.

Two people arguing under these conditions are doing something stranger than disagreeing. One may be running on the felt certainty, the sensation already arrived and settled, while the other is offering evidence and expecting it to move something. It doesn’t move, because evidence acts on a different system from the one that produced the feeling. You can watch this happen and feel the futility of it as it goes: the words are landing on a conviction that words didn’t build and can’t take apart. It’s like trying to reason someone out of a taste, or out of nausea.

The tempting story is that the other side runs on emotion while I run on evidence, but it doesn’t work that way.

I have the same feeling of knowing they do, and from the inside mine feels exactly the same as theirs. The difference that counts is not who has the feeling. It’s whether the feeling has been left connected to anything that could correct it. Have I named, in advance, the fact that would change my mind? Is there a door open through which a piece of evidence could walk in and move me? If there is, my certainty is disciplined. If there isn’t, I’m doing the same thing I accuse the other person of, but I’m just doing it with more footnotes.

That test is a discipline; it’s not innate. It’s available to anyone, on any question, and it’s refused by very clever people every day, because keeping the door open is uncomfortable.

The commonest way of refusing it has an old name: contempt prior to investigation, deciding that a question is beneath looking at so that you never have to look. It’s a very human and efficient move because sneering at a question takes far less effort than studying it. Everyone reaches for it on the questions where they already feel certain. A real door means being wrong sometimes, in public, about things you were sure of. Most of the certainty in the world is bought by quietly nailing that door shut and then admiring how solid the wall feels.

The Shakespeare authorship question is, among other things, a clean example of all this. Set aside for a moment who wrote the plays. The argument about it is a standing demonstration of the difference between a certainty that grades itself against the documentary record and a certainty that has stopped needing to. When someone insists there is no good reason to doubt the man from Stratford and produces no lifetime document tying Shakspere to the writing, the confidence is real and the connection to evidence is not; the two have come unhooked, and a feeling is being offered in place of the missing record. The reply with any hope of success is not a louder counter-certainty. It’s the open door: name the document that would settle it, say plainly that it would settle it, and mean it.

So the site asks one thing of anyone who disagrees. Feel as certain as you like; certainty is not the problem, and no argument can talk you out of it anyway. Just keep that certainty tied to evidence that could change it, because a certainty that nothing could ever move is worth nothing, however strong it feels.

Related: The Self-Check turns this into seven questions to put to your own case before anyone else’s; What Would Settle It names the document that would end the argument here.