The Press Secretary
There is a part of you that does things, and a part of you that explains them. They are not the same part, and they do not have the same information.
The part that decides is older. It runs on signals you cannot read directly — interoceptive feedback from the gut, threat assessments running below conscious awareness, predictive models laid down before you had language. By the time a decision surfaces, it has already been made. What rises into awareness is not the choosing. It is the result of the choosing, plus a story about why.
The story is generated by a different system. Roger Sperry and Michael Gazzaniga's split-brain research, beginning in the 1960s, made this visible in a way that's difficult to argue with. Patients whose corpus callosum had been severed could be given an instruction through the right hemisphere — get up, walk — and they would comply. Ask the left hemisphere, where language lives, why they had done it, and an answer would arrive immediately. I wanted a drink. I needed to stretch. The answer was confident, coherent, and entirely fabricated. The hemisphere that spoke had no access to the hemisphere that acted. So it did what it always does. It generated a plausible account.
This system did not switch off in people whose brains are intact. It runs in all of us, all the time. The conscious mind is not the executive. It is the press secretary — walking out to the podium after the decision has been made, working from a partial briefing, and explaining the administration's position with a confidence that bears no necessary relationship to the truth.
Where Truth Lives
Truth, in any rigorous sense, is not something the inner observer can produce on its own. It is correspondence between a model and something outside the model. The model is generated internally. The thing it must correspond to is not.
This is why physics has gone further than introspection. Not because physicists are wiser, but because they built the discipline to push against the model from the outside. Mathematics as a language with less slack in it than ordinary words. Instruments that extend the senses past their evolved range. The requirement that a theory predict the next measurement, not merely accommodate the last one. When prediction fails, the theory is wrong, regardless of how elegant it was or how many people believed it.
Even there, the answers are provisional. Newton's mechanics described the world accurately enough to navigate ships, build bridges, and send objects to other planets — and then Mercury's orbit didn't quite fit, and the Michelson-Morley experiment found no ether, and a more accurate model became necessary. Newton wasn't wrong, exactly. He was a model with a domain of validity that no one had mapped the edges of. Relativity extended the domain. Quantum mechanics covered another region. Neither is final. They are maps with known good coverage, and the work of finding where they break is ongoing.
The honest stance, even in the discipline that has tried hardest, is asymptotic. Closer, not arrived.
Inside, the situation is harder. There is no Mercury's orbit for your own motivations. The thing you are trying to model is the thing doing the modeling, using equipment built by evolution for survival rather than accuracy. The instruments are not independent. The data is not clean. The system being measured changes in response to being measured.
This does not mean truth is unreachable. It means it is expensive, and the cost is not always worth paying.
The Second Question
When truth is out of reach, or available but too costly to chase, there is a second question that still functions. Not is this story true, but what does this story make me do.
Narratives are not decorative. They generate behavior. The story you tell yourself about your partner shapes whether you reach for them or pull back. The story about your body shapes whether you move it or numb it. The story about a hard day shapes whether you end it on a trail or in a bottle. Whatever account is running in the background is also a program, and the program runs.
Two stories can be equally unverifiable and lead to completely different lives. I need to take the edge off tonight and I need to move my body tonight are both post-hoc accounts of the same underlying state — a nervous system that has accumulated load and is asking for regulation. Neither is true in any deep sense. Both will work, in the sense that both will change how the body feels. But one of them compounds toward a person who can handle more load next week. The other one compounds toward a person who can handle less.
When you cannot verify a story, you can still evaluate it. Run it forward. A week, a year, a decade. Watch where it lands you. The criterion is not whether the story is accurate but whether the person at the end of it is someone you would want to be.
This is harder than it sounds, because the older, more practiced story usually feels more honest. The deepest groove is not authenticity. It is just the deepest groove. A nervous system that has been managing itself through avoidance for thirty years will experience avoidance as natural and the alternative as forced. That sensation is not evidence about which story is true. It is evidence about which one is more wired in.
The choice, when it can be made consciously, is between two costs. The story that compounds well usually has a higher cost in the moment — the discomfort of the trail, the awkwardness of the conversation, the loneliness of not reaching for the phone. The story that compounds poorly usually has a lower cost in the moment and a higher cost over time. Both are real. The math only becomes clear if you let yourself look at the whole tape.
Authoring
Run the loop a few more times. A story generates an action. The action generates a consequence — but it also generates a self. The body that ran the trail is not the same body that didn't. The nervous system that stayed in the difficult conversation has wiring the one that walked away does not. Each iteration of the story lays down the person who runs it next time.
This is why the question of which story to tell is not cosmetic. You are not choosing between explanations. You are choosing between people you will become. The story is the seed. The actions are the growth medium. The person at the end of a year is not the person who started, plus some outcomes. They are the cumulative result of every micro-decision the story made on their behalf.
Most people experience this passively. The story they inherited — from parents, from culture, from the deepest grooves of their own avoidance — keeps authoring them, and they call the result who I am. It is more accurate to call it who the story made.
The Same Machinery
It is worth noticing that large language models hallucinate in exactly the way human introspection does. Both are prediction engines, generating the most plausible next token given context. When the context is tightly constrained by data, both stay close to truth. When it isn't, both confabulate fluently — same confident tone, same coherent surface, no internal flag marking the parts that were invented.
The difference is the body. A human has hunger, pain, exhaustion, the wall they walked into. Ground truth gets injected whether they want it or not. The model can be wrong about the world for as long as the conversation lasts; the body can only be wrong about the world until it falls down.
The mechanism is fragile in an interesting way. The body keeps you honest only when the world is talking back to it. Put a person in an anechoic chamber — a room engineered to absorb essentially all sound — and within about twenty minutes most people find it unbearable. Without ambient sound, the nervous system loses one of its constant low-level references for where it is in space, and starts amplifying internal noise instead. Heartbeat, blood flow, tinnitus, joints. Some people hallucinate. The prediction engine, deprived of correction, begins running unconstrained. The body's vote gets quieter and the spokesperson gets louder. Solitary confinement does the same thing on a longer timescale, with worse consequences.
So embodiment is not a guarantee. It is a feedback loop, and the loop requires a world to push against. Cut the signal and even the body starts to drift.
Two Levels, One Practice
There are two moves available, and they work together.
The first is loosening your grip on the story you are currently telling. Recognizing that the explanation in your head is being generated by a system that does not have full access to the decision it is explaining. Holding your own narratives the way a physicist holds a model — useful, provisional, accurate within some domain, almost certainly wrong at the edges. This does not require you to stop telling stories. It requires you to stop confusing them with the territory.
The second is choosing, where you can, the stories that generate the actions you would endorse on reflection. Not because those stories are more true, but because action has consequences and consequences compound. The story that puts you on the trail tonight wins not on epistemic grounds but on the grounds that, six months from now, you will be a person who can run further.
Neither move is dramatic. Neither produces certainty. Together they produce something more useful than certainty, which is the capacity to keep adjusting — to hold a story loosely enough that you can update it when the evidence comes in, and firmly enough that it still gets you out the door.
The press secretary does not stop talking. You stop believing every briefing.