How accurate is your memory of your own life?

Willem Wagenaar spent six years answering a question most people only ask mid-argument: how accurate is your memory of your own life? He logged 2,402 moments from his own life, then tried to reconstruct them using nothing but the notes he had written at the time. It remains the most detailed study anyone has run on how much a memory holds up when nothing else confirms it.

What exactly did he record, and how was he tested?

For six years, Wagenaar logged every distinguishable event in his life on a card with four fields — who, what, where, and when. He added a critical detail that only someone who truly remembered the event could answer. He rated each one for how often that kind of thing happened, how emotional it was, and how pleasant. He never reread a single card before testing himself.

In total he recorded 2,402 events. Of those, 1,605 came from the four-year main experiment. The other 797 were split across a control year before and one after, to check that his memory hadn’t changed quality while he was logging. At recall, each event appeared in a booklet with its cues shown in a different order — 24 possible sequences in all. He had to reconstruct the rest from scratch. He was the only subject: the researcher and the person being studied were the same man. He was 37 when the recording began and 43 at his last recall attempt.

What did he compare his results against?

Wagenaar checked his own performance against two control years: one before the main experiment, one after. He wanted to rule out that his memory had gotten better or worse while he was logging. Both control years matched the four years in between. That suggests neither his memory’s quality nor the kind of events he recorded changed over time.

He also placed his numbers next to the two precedents that existed. Marigold Linton’s self-study, presented in 1975 and published in 1978, tracked events over seven years, but measured only dating accuracy. Robert White’s 1982 study used a simpler method: one sentence per event. Wagenaar cites both as reference points, not as experimental controls. They’re two other single-person cases, not a sample large enough to average.

How steep was the decline, and what held up?

Recall dropped steadily. About 70% of questions were answered correctly at six months. By five years, the figure had fallen to somewhere between 29% and 33%. The full curve fits a simple power function (y = 0.54x⁻⁰·³⁶, correlation of .99). It’s steep early on, then slower and slower. It isn’t a straight line toward forgetting.

The single most useful cue for pulling up an event was knowing what happened. Given the what, he could reconstruct who was there 62% of the time, and where it happened 58% of the time. Knowing when it happened, by contrast, was almost useless as a cue: it led him to the right person only 2% of the time. That’s a strange asymmetry. A date is easy to reproduce once you’ve found the event some other way, but nearly useless as a starting point for finding it. Stacking two or three cues together helped more than adding up their separate effects would predict. The exception was the date — combining it with another cue sometimes made the pair perform worse, not better.

Pleasant events were remembered better than unpleasant ones: 53% versus 34%, with neutral events in between at 36%. The gap isn’t explained by frequency — pleasant events were, in fact, far more common in his life than unpleasant ones. And the finding Wagenaar underlines in his own discussion: not one event came back as fully forgotten. Even the ones his four-cue system couldn’t retrieve told a different story later. When he tracked down other people who had been there, the events turned out to have enough surviving detail to be recognized.

Practice didn’t help either. Wagenaar compared his performance in the first half of the recording period against the second half. He was checking whether logging thousands of events had made him better at recalling them. It hadn’t. Recording and retrieving memories, month after month, didn’t train the skill the way a muscle trains.

What this study doesn’t say

The biggest limit is one Wagenaar states himself: he was his only subject. There’s no guarantee his memory works the way anyone else’s does. His age and his training as a researcher used to logging with precision are real variables. So is the discipline it took to keep this up for six straight years. This study can’t average any of them against anyone else.

It also doesn’t measure what happens to memory when you write about something difficult. Whether the act of writing changes how much gets remembered is a different question, with different science behind it. And the claim that “no event was fully forgotten” depends on how forgetting gets defined. Wagenaar defined it as being unable to retrieve information given a limited set of cues, not as the trace being gone entirely. About 20% of events came back unrecoverable under his own four-question system. It was outside help that rescued them afterward, not the system on its own.

What holds up

What these numbers support, with those conditions written in plain sight, is this. A whole life’s memory doesn’t work like a filing cabinet searched by date. Wagenaar backs that up with two findings of his own. More than 80% of the questions he couldn’t answer, he left entirely blank, instead of guessing a recent date and getting it wrong. That’s exactly what a chronological search would do. And his dating errors landed just as often too far in the past as too far in the future. An archive searched backward from today should skew almost entirely toward recent guesses; his didn’t. Search by calendar, and neither result makes sense. What organizes recall is content — what happened — not when.

This study answers a different question than rereading your own journal years later does. That piece is about how the impression of a whole stretch of time changes when you read it in one sitting. It isn’t about how much of it you can recover without any record at all.

It’s different again from why a year-end review always sounds kinder than the year actually felt. That’s a bias about how good the past feels. It isn’t about how much of the past is still there to recall.

With or without a notebook, the exact date is the first thing to go. What happened is the last.