The Peak-End Rule: Why You Remember the Worst Moment and the Last One

If you ask someone to rate a painful or pleasant experience after it's over, they won't average how they felt moment to moment. They'll report something close to the average of the worst (or best) moment and the last moment, and they'll barely register how long it lasted. That's the peak-end rule.

Origin

Daniel Kahneman, working with Barbara Fredrickson and colleagues, ran a cold-pressor experiment in the early 1990s. Subjects held one hand in 14°C water for 60 seconds, then repeated the trial with the other hand: the same 60 seconds, followed by an extra 30 seconds where the water was warmed slightly to 15°C. Objectively the second trial involved more total discomfort. Subjects still preferred to repeat it, because it ended on a milder note. Kahneman called the near-total disregard for length "duration neglect."

The finding scaled up in a 1996 study with Donald Redelmeier: 682 patients undergoing colonoscopies were randomly assigned to a standard procedure or one where the scope was left in place briefly at the end, adding mild discomfort but no new pain spikes. Patients in the extended group rated the procedure as less unpleasant overall, and were more likely to return for a follow-up screening years later. A longer, technically worse procedure was remembered as better, because it ended more gently than it peaked.

What the evidence actually supports

The cold-pressor and colonoscopy studies are tight, well-controlled, and have replicated reasonably well: short, bounded experiences with a clear peak and a clear end. That's the domain where the peak-end rule is on solid ground.

It gets shakier once you stretch it. "How was your week" or "how was this year" aren't bounded fifteen-minute intervals with one discomfort curve. They're bundles of many separate episodes, each with its own peak and end, and it's not obvious the same simple averaging rule survives that jump. Popular write-ups often apply the peak-end rule to long stretches of time as if the mechanism were proven at that scale. It's a reasonable hypothesis, not an established result. Worth knowing the difference before you lean on it.

What it means for keeping a record

The practical risk isn't abstract. If you sit down to write a daily note at 11pm after a rough final hour, that entry will skew toward the rough hour, not the day as a whole. Duration neglect doesn't care that the other fifteen waking hours were fine. The same distortion hits an end-of-project retro written the day after a chaotic launch: the launch chaos, being both the peak and the end, crowds out the calmer months of steady work that preceded it. You'd walk away thinking the whole project was stressful when most of it wasn't.

This is a straightforward argument for how to journal as you go rather than reconstructing a period from memory at its close. A running log, or notes captured near the time something actually happened, isn't subject to whichever moment happens to be loudest in your head at write-up time. It's also part of why a weekly review works better when it pulls from notes taken throughout the week instead of being written cold from memory on Friday afternoon, when Friday afternoon's mood is doing all the talking.

The fix isn't complicated: capture moments closer to when they happen, and treat any summary written well after the fact as a compressed, peak-and-end-weighted version of events rather than an accurate average. This sits next to other memory quirks worth knowing about, like the serial position effect, where what you remember from a list depends on where in the sequence it fell, and the Zeigarnik effect, where unfinished tasks linger in memory more than finished ones. All three point the same direction: raw memory is a bad archivist, and the record you write in the moment is more trustworthy than the one you reconstruct later.

More in Memory & Focus

75 more notes on this branch.