Introspection Is a Press Secretary, Not a President

Of all the things people take themselves to know, the contents of their own minds seem the least open to doubt. You might be wrong about the weather or the date or what your colleague meant, but surely not about why you chose the blue one. That knowledge feels direct, unmediated, and authoritative. You were there. You did it. You know why.

This confidence is misplaced, and the experiments demonstrating so are among the most unsettling results in psychology. Introspection does not report on the process that produced your decision. It produces a plausible account after the decision, using much the same guesswork it would use to explain a stranger's behaviour — and it does so without any sense that it is guessing.

The Report That Was Simply Invented

The clearest demonstrations come from split-brain patients, in whom the corpus callosum connecting the hemispheres has been severed to control epilepsy. The two hemispheres continue functioning but can no longer exchange information directly.

Michael Gazzaniga's experiments exploited this. Present an instruction — "walk" — to the right hemisphere only, and the patient stands up and heads for the door. Now ask why. Speech is generated by the left hemisphere, which never received the instruction and has no access to the real cause. It does not say "I don't know." It says: "I'm going to get a Coke."

The explanation is fluent, immediate, and delivered with total confidence. It is also entirely fabricated. The left hemisphere has observed the body rising and constructed a reason, exactly as it would for anyone else's behaviour, and then presented that construction as a memory of its own intention.

Gazzaniga called this system the interpreter. Its job is not to know why you act. Its job is to produce a story about why you acted, continuously, whether or not the information required is available.

Choice Blindness

The obvious objection is that split-brain patients are an unusual population. So consider what happens in ordinary intact brains.

In Johansson and Hall's experiments, subjects were shown two photographs of faces and asked which they found more attractive. The chosen photograph was then handed to them, and they were asked to explain the choice. Using sleight of hand, the experimenters sometimes handed over the photograph the subject had rejected.

In roughly three-quarters of these swapped trials, subjects failed to notice. They then explained, at length, why they had picked the face they had in fact just turned down. "I liked her earrings." "She looks more approachable." The reasons were confident, specific, and about a choice nobody had made.

The same result has been obtained with tastes of jam, with political opinions on a questionnaire, and with financial decisions. The explanation machinery does not check whether it is explaining a real choice. It explains whatever it is presented with.

Nisbett and Wilson

The foundational paper here is Nisbett and Wilson's 1977 review, which asked directly how much access people have to their own mental processes and concluded: often none at all.

In one study, shoppers were shown four identical pairs of stockings and asked to choose. There was a strong position effect — the rightmost pair was chosen about four times as often as the leftmost. Asked why, not a single subject mentioned position. They cited knit, sheerness, weave — properties of items that were literally identical. When the position effect was suggested to them, they denied it, sometimes with irritation.

That last detail is the important one. It is not merely that the real cause was unknown. It is that a confident and specific alternative was supplied instead, and defended.

Why It Feels Like Knowledge

If introspection is this unreliable, why does it feel like the most certain thing we have?

Because there is no signal to tell you otherwise. When your eyes are working badly you get blur — an error indication. Introspection has no equivalent. The interpreter produces an account, and that account arrives labelled as memory of your own reasoning, indistinguishable in feel from cases where the reasoning genuinely was accessible.

Consider the design constraints. The processes that actually generate behaviour are distributed across enormous numbers of neurons doing things with no verbal form at all. There is no channel by which that could be reported. What the verbal system can do is take the output — the action, the preference, the feeling — and construct an account consistent with what it knows about the person it belongs to. That account is often right, because we do know a good deal about ourselves in general. But it is arrived at by inference, not by observation, and when the inference is wrong nothing flags it.

Which is why "press secretary" is the right analogy. The press secretary was not in the room. They receive the decision after it is made, and their function is to explain it in terms that make the administration look coherent. They are frequently accurate — and they are never reporting from direct knowledge, and never say so.

What This Costs

This is not an academic curiosity. A great deal rests on introspective testimony.

Legal systems treat a defendant's account of their own motive as evidence. Job interviews treat a candidate's explanation of their choices as informative. Therapeutic traditions have been built on the premise that sufficiently careful introspection recovers the true causes of behaviour. Every argument that begins "I know my own mind" leans on the same assumption.

And it underwrites the folk theory of the self: that there is someone inside, in possession of the reasons, issuing the decisions. The evidence points elsewhere. The decisions are produced by machinery you have no access to; the "someone inside" is the department that writes them up afterwards.

There is a further irony. The introspective sense of a unified self in charge is the very thing produced by the system least qualified to report on the matter. The witness testifying to the president's existence is the press secretary.

Living With It

The reasonable response is not to distrust everything you think about yourself. It is to demote the feeling of certainty, which is the part carrying no information.

Practically, this means preferring evidence to testimony where the two can be compared. What you actually did, repeatedly, over time, is better data about your preferences than your account of why you did it. It means taking behavioural findings about human beings seriously even when they do not match your introspective sense of yourself — because your introspective sense is generated by a system demonstrably willing to fabricate. And it means treating other people's accounts of their motives with the same measured scepticism, which is more charitable than it sounds: most confabulation is not lying. The person genuinely believes the story. They were told it by the same machinery that told you yours.

This is another entry in the long demotion. We lost the centre of the cosmos, then our separateness from the animals, then the immaterial mind. This one takes the last redoubt — the assumption that whatever else we are wrong about, we at least have privileged access to ourselves. We do not. We are, on the inside, in roughly the same position as an outside observer with unusually good access to the raw footage and unusually strong motivation to conclude that it all makes sense.