Nine people said it was good. One person said the middle section did not work. Three weeks later you can produce that sentence verbatim, including the pause before it, and the nine have merged into an undifferentiated warm blur that carries no weight at all.
This is so consistent across people that it is better understood as a design feature than as a fault, and the first useful move is to stop treating it as evidence of low self-esteem. Plenty of confident people do exactly this. What separates them from the people it wrecks is not how much the remark stung at the time; it is what happened in the following hour.
What is usually going on
Negative information is weighted more heavily, by default, in nearly everyone. The asymmetry shows up across an unusually wide range of research — impressions, learning, memory, decisions — and the plain reason is that the costs of the two errors were never symmetrical. Missing a compliment costs nothing. Missing a warning was historically expensive. A system built on that arithmetic produces exactly the memory you have.
Praise is vague and criticism is specific, and memory prefers specifics. That was great has no content to encode. The second half loses the thread is a pointer to a location, and it comes with an image of the person saying it. Much of what feels like selective retention is simply the difference between remembering an atmosphere and remembering a sentence.
You discount praise for a reason that is not irrational. People say encouraging things for social reasons constantly, and you know this because you do it. Criticism is costly to deliver, so it reads as more likely to be sincere. The inference is sound; the error is applying it universally, including to the people whose praise is as considered as their criticism.
It confirmed something you already suspected. The remark that stays is rarely the harshest one. It is the one that matched a private doubt, which is why other people's sharper comments slide off and one mild observation embeds itself for a decade. The staying power belongs to the doubt, not to the remark.
It was heard as a verdict rather than as information. This is the part that decides how long it lasts. This draft has a problem is a fact about a draft, and it is actionable within the hour. I am not good at this is a ruling about a person, and there is nothing to do with it except carry it. The words said were probably the first kind. The translation happened on your end, automatically, and it is the only part of this that is really adjustable.
And the honest one: you would not actually want it gone. A person who genuinely weighted praise and criticism equally would be worse at their job and harder to work with. The sensitivity that makes one comment expensive is the same sensitivity that catches the problem before it ships. The target is not indifference. It is a shorter half-life.
What to do
- Convert it into a task within a day. Write the criticism as a sentence that starts with a verb — rewrite the middle section — and put it on a list. A remark with an action attached stops being a statement about you and becomes an item, and items get closed.
- Ask for specific praise, which people will happily give and almost never volunteer. What specifically worked produces something with as much content as the criticism had, and therefore something with a chance of being remembered.
- Keep the receipts, in writing, because your memory will not. A file of actual sentences people wrote is not a self-esteem exercise; it is a correction for a known measurement bias in your own recall.
- Watch the translation as it happens. The moment to catch is the one where a note about a thing becomes a statement about a person, and the question that interrupts it is simply: what were the actual words, and did they contain me as a subject.
- Stop trying to feel differently about it, and manage the hour afterwards instead. The sting is not optional and arguing with it doubles the cost. What is optional is whether that hour is spent reconstructing the conversation or doing the one edit it implied.
What it depends on
How long a remark lasts depends on what you believe it measured. Under the assumption that ability is a fixed quantity, every piece of feedback is a reading of that quantity and therefore permanent news; under the assumption that it is built, the same sentence is a report on a method, and methods get changed on Thursday. The words are identical. The half-life differs by weeks.
That belief is not evenly distributed across your life — most people take feedback cheerfully in domains that do not define them and badly in the one that does — which is why the facet breakdown is more informative than the overall score. The wider version of that stance covers what it looks like elsewhere, the comparison habit is the same machinery pointed at other people's results, and if the remark has stopped being about the work entirely and settled into a statement about what you are, that is shame rather than feedback and it needs different handling.