Flashbulb memories are lying to you. Your data is not.
On January 28, 1986, the space shuttle Challenger broke apart 73 seconds after launch. The next morning, the psychologist Ulric Neisser asked 106 students in his introductory class at Emory to write down how they had heard the news: where they were, what they were doing, who had told them, how they had felt.
Two and a half years later, 44 of those students came back. Neisser and his student Nicole Harsch asked the same questions again, then handed each participant the original account in their own handwriting.
The discrepancies were extreme. One student had written in 1986 that she was in her religion class when a classmate announced the explosion. In 1988 she remembered being in her dorm room, watching the launch live on television with her roommate. She was certain of the second version. Shown the first, she looked at her own handwriting and said, “I still think of it as the other way around.”
That sentence is the problem with flashbulb memories in one line. She did not disbelieve the 1986 document. She just could not align it with what she now remembered. Her confidence in the new version was not moved by finding out it was wrong.
The original claim
The phrase “flashbulb memory” was coined by the Harvard psychologists Roger Brown and James Kulik in a 1977 paper in Cognition. They studied what Americans remembered about hearing that President Kennedy had been shot, that Martin Luther King Jr. had been assassinated, that someone close to them had died. Their hypothesis was that consequential, emotionally charged events triggered a distinct memory mechanism: something like a neural snapshot that recorded the surrounding context with unusual fidelity. Respondents could describe the room they were standing in, name the person who had delivered the news, recall the quality of the light, even decades after the event.
Brown and Kulik called this a “Now Print!” mechanism, an evolutionarily old system that flagged important moments and kept them. The paper was hugely influential. It gave academic shape to an intuition everyone already had: that the brain holds some moments in full and lets the rest drain away.
The intuition was half right. The moments are treated specially. The special treatment is not what anyone thought.
What confidence is, and what it is not
The crucial finding arrived in 2003. On September 12, 2001, the day after the World Trade Center attacks, Jennifer Talarico and David Rubin at Duke recruited 54 students and asked them to record their memory of hearing about the attack, alongside their memory of an ordinary event from the recent past. They tested the same students again at 1, 6, or 32 weeks.
At any delay, the flashbulb memories and the ordinary memories were about equally consistent with the original record. Both drifted at roughly the same rate. What differed was confidence. Participants kept rating their 9/11 memories as vivid, precise, and trustworthy. Their confidence in the ordinary memory tracked its actual drift and declined. Their confidence in the flashbulb memory did not.
The paper’s title is the whole finding: “Confidence, not consistency, characterizes flashbulb memories.”
This is a significant inversion of the original claim. A flashbulb memory is not a more accurate memory. It is a more believed memory. The vividness is not evidence of fidelity. It is a side effect of emotional intensity at encoding, and it stays high even while the content quietly drifts underneath it.
Ten years, and a different kind of finding
In 2015, William Hirst and a large consortium of researchers published a ten-year follow-up to their own 9/11 memory study in the Journal of Experimental Psychology: General. They had surveyed participants within a week of the attacks, then again at 11 months, 25 months, and 119 months. This let them watch memories not just decay but reshape.
The study produced two findings worth separating.
First, the forgetting happened early. Most of the loss of consistency occurred in the first year. The curves then flattened. The memory at year ten was not much worse than the memory at year one. It had stabilized at a certain distance from the truth.
Second, the stabilization was not a return to accuracy. When a participant produced an inconsistent memory at the 11-month mark, they tended to produce the same inconsistent memory at year three, and again at year ten. Inaccurate facts about the event itself, which people continued to encounter through the news, were often corrected. Inaccurate personal memories, the ones about where you were and what you felt, were preserved. The wrong version became the version.
Each retelling reinforced the last. Confidence rose with repetition. The memory that felt most certain was often the one that had diverged the furthest from the original record.
Your ordinary flashbulbs
Most people’s flashbulb memories are not public. They are the call from the hospital. The first date. The morning the funding came through. The afternoon the company folded. The text message read at the bus stop. The apartment walked into and signed for the same week.
These memories carry the same signature as the 9/11 memories in the lab: high confidence, vivid sensory content, a certainty that this one is preserved exactly. They also carry the same risk. Every time the story of how you met has been told, the telling has shifted the memory a little. The version you are confident about today is not the version you held at the time. You have no way of knowing how far it has moved, because the only record of the original is the same memory that did the editing.
The drift is not a failure of effort or care. It is a structural feature of how human memory operates. Each act of recall is reconstructive rather than reproductive, informed by what you now know, who you have become, and what you have told other people about the event. Neisser’s Challenger participants were not careless. They were doing exactly what brains do with their most treasured memories.
What does not revise itself
The thing that separates an external record from an internal memory is not detail. It is that the external record does not change when you change.
A calendar entry for that Thursday does not acquire a different meeting title when you later decide what the meeting meant. The commit pushed at 4:47 PM does not gain or lose significance depending on whether the project eventually shipped. The Slack message sent to your co-founder is phrased the way it was phrased, not the way you now wish you had phrased it. The check-in at a restaurant names the restaurant, not the one you later misremember being at.
None of this is a richer form of memory. It is a different kind of evidence. Cross-checking a vivid recollection against a plain record is what lets you notice that you were sure a conversation happened on a Tuesday when the calendar shows it was a Wednesday, that the meeting you remember as a breakthrough came two weeks after the real turn, that the person you first met at the conference was actually introduced to you at dinner the night before. That kind of check is rarely available, because most people do not keep contemporaneous records of an ordinary week. The week passes, the memory forms, and nothing ever tests it.
The diary as a ledger your brain cannot edit
deariary builds a diary entry each day from the traces your tools already produce: the meetings on your calendar, the messages you sent, the commits you pushed, the places you checked in, the music you listened to. The entry on the day it is written is unremarkable. The value arrives later, on the day you find yourself telling a story about that period and half-wondering whether that is really how it went.
You can keep the vivid memory. The research is clear that the vividness is doing something for you: emotional intensity at encoding produces a feeling of specialness, and that feeling is worth having. What you cannot do, without an outside record, is distinguish the parts that survived from the parts that have been rewritten on each retelling.
Brown and Kulik proposed a mechanism that printed the important moments and left the rest. The research that followed showed that brains do no such thing. They keep a bright impression and a drifting content, and the impression does not know when the content is wrong.
A diary assembled from data is not a better memory than your own. It is a record your memory cannot reach back into and edit.
