There is an argument in your past that you lost twice. Once in the room, and once again every time it resurfaces, because the other person remembers it differently and will not budge. They swear you said the cruel thing first. You know, with the full weight of your body, that they did. The unsettling part is not that one of you is lying. It is that neither of you is. Two honest people walked out of the same kitchen carrying two different nights, and both of them feel like the truth.
This happens with the good days too. A friend brings up the evening you both call the best of that summer, and the version they tell has details you would have sworn belonged to a different year. You laugh it off. But somewhere underneath, a small question opens: if we can't agree on what happened, what exactly am I remembering?
Your brain keeps two copies, and one of them is cheap
The most useful answer comes from a framework in cognitive psychology called fuzzy-trace theory, developed over decades by Charles Brainerd and Valerie Reyna. Its central claim is simple and a little humbling. When you experience something, your mind does not file a single recording. It lays down two kinds of trace at once.
The first is verbatim: the surface of the event. The exact words, the order they came in, the precise phrasing, the tone of the third sentence. The second is gist: the meaning. What it was about, what it felt like, what it implied about you and the other person.
These two traces have very different shelf lives. Verbatim memory is expensive to hold and fades quickly, often within hours or days. Gist is cheap, durable, and sticky. Weeks later, when you reach for the argument, the verbatim copy has mostly dissolved, so your mind does what it always does with a gap: it rebuilds. It takes the gist that survived, 'they made me feel small', and reconstructs plausible words to fit it. The reconstruction arrives with the confidence of a recording, because from the inside, remembering and reconstructing feel identical.
Now put two people in the same room. Each walks out with a different gist. Each loses their verbatim at roughly the same rate. Each rebuilds the dialogue from their own meaning. Of course the scripts don't match. They were never written from the same outline.
Why your gist and their gist were never the same
If gist is meaning, the next question is why two people extract different meaning from the same ten minutes. Part of the answer is a well-documented pattern called egocentric bias.
In a classic 1979 series of studies, Michael Ross and Fiore Sicoly asked married couples to estimate what percentage of household tasks each of them did. When you added the two partners' estimates together, they routinely summed to more than one hundred percent. Nobody was lying. Each person simply had far richer access to their own contributions than to their partner's, because you are present for every dish you wash and only some of the ones they do.
The same asymmetry runs through a conversation. You have direct access to your own intentions. You know you said the sharp thing because you were frightened, or tired, or trying to end it. The other person has no access to your intention at all. They only have your effect. So the gist you encode is 'I was defending myself', and the gist they encode is 'I was attacked'. Both are accurate descriptions of the same event, from the only vantage point each of you was given.
This is why 'what did you actually say' is such a hopeless question months later. You are asking two people to produce verbatim from gist, and their gists were pointed in opposite directions from the first minute.
The edits that happen after the fact
There is a third force, and it is the one that quietly makes things worse with time. Elizabeth Loftus and her colleagues spent decades showing that memory is porous to information that arrives after an event. In her famous early work with John Palmer, people who watched the same footage of a car collision gave different speed estimates depending on whether the question used the word 'smashed' or 'hit', and the 'smashed' group was more likely to later recall broken glass that had never been there.
This is the misinformation effect, and it does not require a manipulative questioner. A retelling is enough. Every time you describe the argument to a friend, you phrase it a certain way, your friend reacts, and that phrasing and that reaction get folded back into the memory. By the tenth telling, the memory contains the story's edits. The other person has been doing the same thing in a different kitchen with different friends. The two accounts do not merely fail to converge. They drift apart on purpose, each one sanded smooth by its own audience.
This isn't only about fights
It is tempting to treat this as a problem of conflict. It isn't. The same machinery takes your best days.
Someone once said something to you that you decided to keep forever. A compliment that landed in a place nothing else had reached. You remember that it happened and roughly what it meant. Try now to recall the actual sentence. Most people find the words have gone soft, replaced with a paraphrase that sounds like them rather than like the person who said it. The gist survived. The verbatim, the thing that made it theirs, is gone.
That is the quiet cost of running on gist alone. You keep the meanings of your life and lose the voices. And because gist is shaped by what you already believed, a life remembered only in gist slowly becomes a life that confirms you, rather than one that surprises you.
The practical conclusion is almost embarrassingly concrete. Verbatim memory has one reliable storage medium, and it is not your head. It is the page, written close to the event, before the reconstruction begins.
Your next moves
- Tonight, write one exchange in quotation marks. Pick a single moment from today, good or hard, and record who said what in direct quotes, as close to the actual words as you can get. Do it within a few hours. Tomorrow is already too late for most verbatim.
- Keep the words and the meaning in separate lines. Under the quote, write a second line that starts with 'What I think it meant:'. Now you have a verbatim trace and a gist trace stored side by side, so your future self can see where the meaning came from and whether it still holds.
- Change the question you ask in a disagreement. When you and someone you love remember an event differently, stop litigating the words. Ask 'What did it mean to you?' instead. You are trading an argument about two lost recordings for a conversation about two real experiences.
- Run the hundred-percent check on one shared thing. Pick something you and a partner or housemate both do. Each of you privately writes your share as a percentage, then compare. If the total is over a hundred, you have just watched egocentric bias in your own handwriting, and it is much harder to resent someone after that.
- Date every entry and never edit an old one. If you want to add what you understand now, add it below with today's date. The original must stay as it was, because its whole value is that it was written before the retellings started.
Where a daily record earns its place
All of this points at the same small discipline: catch the day before the day becomes a story. Lore is built around exactly that. Every day tells a story, but the one worth keeping is the one you wrote the night it happened, with the real words still in it, before anyone, including you, had the chance to revise. You don't need an app to do this. A notebook and a stubborn habit will work. But if you want somewhere that makes tonight's version easy to write and impossible to quietly overwrite, you can start at https://lore.lumenlabs.works.