Someone asks how your week was, and you do the thing everyone does: you scan backward, wait for a verdict to surface, and say it out loud. Honestly, kind of rough. It feels like a fair summary. It almost never is.
What surfaces when you scan a week is not an average of the hours you actually lived. Memory doesn't keep the hours. It keeps a highlight reel — and the reel is edited by a rule so consistent that psychologists gave it a name: the peak-end rule. When you evaluate a past experience, your judgment is assembled almost entirely from two moments: the most emotionally intense one, and the final one. Everything else — including how long any of it lasted — barely gets a vote.
Understanding this one editing rule changes how you interpret your own moods, your own stories, and the confident verdicts you deliver about your past. So it's worth slowing down and watching the editor work.
Two selves, one microphone
Daniel Kahneman, the psychologist whose work with Amos Tversky reshaped how we think about judgment, spent much of his later career on a deceptively simple question: when you say an experience was good or bad, who is talking?
His answer was that two different characters share your life. The experiencing self lives in the present tense. It feels this sip of coffee, this awkward meeting, this walk home — a continuous stream of moments, each one real while it lasts and then gone. The remembering self is the one that keeps score. It writes the story afterward, answers questions like "how was your week?", and — crucially — makes your decisions about what to repeat and what to avoid.
The catch is that the remembering self is not a faithful accountant. It doesn't add up moments. It samples them. And its sampling method is the peak and the end.
The cold water experiment
The cleanest demonstration comes from a 1993 study by Kahneman, Barbara Fredrickson, Charles Schreiber, and Donald Redelmeier. Participants held one hand in painfully cold water — about 14 degrees Celsius — for sixty seconds. In a separate trial, they did the same sixty seconds, but the immersion continued for thirty additional seconds during which the water was warmed slightly: still unpleasant, just less so.
By any sane arithmetic, the second trial is worse. It contains everything the first one does, plus half a minute of additional discomfort. But when participants were asked which trial they'd be willing to repeat, most chose the longer one.
Why? Because it ended better. The remembering self doesn't tally total suffering. It checks how bad the worst moment was and how the thing finished, and files a verdict. A gentler ending retroactively improved the memory of the entire episode — enough that people volunteered for more pain to get it.
Duration neglect: the hours that don't count
The companion finding is stranger still. Redelmeier and Kahneman studied patients undergoing colonoscopies in the era before routine sedation, asking them to rate their pain minute by minute during the procedure. The procedures varied enormously in length. Yet when patients later rated the experience as a whole, duration had almost no influence on their judgment. What predicted the remembered awfulness was, again, the peak moment and the final moments.
Researchers call this duration neglect, a term from Fredrickson and Kahneman's work on emotional episodes: the length of an experience largely drops out of memory's evaluation of it. In a later clinical study, Redelmeier and colleagues even tested the implication directly — extending the end of the procedure with a brief, milder final phase. Patients who got the gentler ending remembered the experience as less unpleasant, and were more likely to return for follow-up screening.
Sit with that. Adding time to a medical procedure made the memory of it better, because the addition softened the ending. The reel doesn't care about runtime. It cares about the trailer.
What this does to your emotional life
The lab studies used cold water and colonoscopies because pain is easy to measure. But the rule doesn't stay in the lab. It's running every time you summarize your own past.
A week with six ordinary-to-pleasant days and one Friday argument becomes "a rough week," because the argument was the peak and it sat near the end. A vacation whose last day dissolved into a missed flight dims the whole trip in memory, even though the first five days were quietly wonderful. A job with rare but intense crises can be remembered as worse than one with constant low-grade grind, because the crises are peaks and the grind — being long, flat, and duration-shaped — is exactly what memory neglects.
And here is the part that matters most: your answers to "how have you been feeling lately?" — the ones you give a partner, a friend, a doctor, a therapist — are peak-end summaries. They're not lies. They're honest reports from an editor who threw away most of the footage. Research on retrospective emotion reports consistently finds gaps between how people rate their days in the moment and how they rate the same days looking back, with the memory version tilted toward peaks and recency.
This is a different distortion from mood-congruent memory, where your current mood biases which memories you can reach. The peak-end edit happens even when you're perfectly calm. It's not a mood problem. It's a structural feature of how episodes get compressed for storage.
Working with the rule instead of against it
You can't switch the editor off. But once you know the editing rule, three moves become available.
Design your endings. Since final moments carry outsized weight, they're the highest-leverage minutes you have. Ending the workday by finishing one small, complete thing — rather than abandoning a task mid-struggle — changes how the whole day files away. The same goes for how an evening ends, how a hard conversation closes, how a trip wraps up. This isn't spin; it's giving the remembering self accurate material where its attention actually falls.
Interrogate the verdict. When "that was a terrible week" surfaces, treat it as a headline, not a finding. Ask what Tuesday afternoon actually felt like. Often you can't answer — which is itself the tell. The verdict was written from two data points, and the honest response to a two-point summary is mild suspicion.
Capture feelings when they happen. This is the only real countermeasure, because the experiencing self testifies exactly once — in the moment — and then hands everything to the editor. A few words written at the time ("3pm: tense before the call, fine after") are testimony the reel can't rewrite. Days later, that contemporaneous record will routinely contradict your memory's summary, and the record is the one telling the truth.
The record the reel can't rewrite
That third move is what Pulse was built for. It takes a few seconds to log what you're feeling while you're feeling it — no essay, no analysis, just the experiencing self getting its testimony down before the editor takes over. Weeks later, you can look back at the actual shape of your days: the ordinary Tuesday afternoons, the mild Wednesdays, the one bad Friday that your memory tried to stretch across the whole week. And because this is testimony about your inner life, it stays yours — your entries live on your device, not on someone's server. Your feelings stay here. If you'd like a truer record than the highlight reel, you can start one at pulse.lumenlabs.works.