The Peak-End Rule: Why Endings Decide What You Remember

Think about the last holiday you took. Do not describe it, just notice what surfaced. Almost certainly you got one bright moment and something near the end, and almost certainly you did not get a proportional summary of all fourteen days.

That is not a flaw in your particular memory. It is a reliable pattern with a name, and once you know it you can use it, because it turns out that endings are doing far more work than they deserve to.

What the peak-end rule says

The peak-end rule holds that our retrospective evaluation of an experience is driven mainly by two data points: the moment of greatest intensity, the peak, and the affective state at the end. Everything in between gets compressed. Most strikingly, how long the experience lasted has very little influence on the remembered rating, an effect Kahneman and colleagues called duration neglect.

This is not a claim that duration does not matter while you are living through it. It matters enormously in the moment. The claim is narrower and weirder: the memory you keep is not a summary of the moments you lived.

The cold water study

The cleanest demonstration is Kahneman, Fredrickson, Schreiber and Redelmeier's cold-water experiment. Participants went through two trials. In the short trial they held a hand in painfully cold water for sixty seconds. In the long trial they held a hand in the same cold water for sixty seconds, and then the water was gradually warmed slightly for another thirty seconds, still unpleasant but less so.

The long trial contained strictly more discomfort. Every second of the short trial was there, plus thirty additional seconds of cold. Yet when participants were asked which trial they would rather repeat, a clear majority chose the longer one. It ended better, so it was remembered as better, so it was preferred.

That is the finding in its most uncomfortable form. Adding pain improved the memory of the experience, because the added pain was milder than the peak and it changed the ending.

The colonoscopy study

Redelmeier and Kahneman ran the same logic in a clinical setting, with patients undergoing colonoscopy in an era when the procedure was considerably less comfortable than it is now. Patients reported their discomfort in real time, minute by minute, which gave the researchers both a moment-by-moment record and a later retrospective rating.

The retrospective rating tracked the peak and the final moments. It did not track the total, and procedures that lasted far longer were not remembered as substantially worse, provided they did not end at a high point of discomfort. In a later trial, deliberately extending the procedure with an added period of low discomfort at the end improved patients' remembered evaluations, and there was evidence it made them more willing to return for follow-up screening.

That last part is why this is not a parlour trick. Whether someone comes back is decided by the remembered experience, not the lived one.

The two selves

Kahneman's framing is that we contain two customers with different interests. The experiencing self lives through moments and cares about how each one feels. The remembering self builds a story afterwards, and it is the remembering self that makes decisions, tells other people about it, and decides whether to repeat the thing.

You cannot satisfy both perfectly, and they will sometimes conflict. A holiday optimised for the remembering self might include an exhausting sunrise hike that produces a photograph and a story. A holiday optimised for the experiencing self might be a week of pleasant nothing that leaves almost no trace.

For learning habits, which are the thing I care about, the remembering self is the one to court. It is the self that decides tomorrow.

How to use this on purpose

End study sessions on a win

The instinct when you hit the hard problem is to keep grinding until you crack it, then stop. That is exactly backwards for habit formation. If you finish at the moment of maximum frustration, that is what your remembering self files, and tomorrow you will find a reason to skip.

Do the hard thing in the middle. Then spend the last two minutes on something you can already do: a quick recall of material you know, a problem one level below your ceiling, a summary in your own words. You are not padding, you are choosing the last frame.

Close one loop before you stop

An unfinished task keeps running in the background and colours the ending badly. If your session must stop mid-topic, spend thirty seconds writing the next step down explicitly, which converts an open loop into a note. The residue that lingers when you switch away without closing anything is worth understanding on its own, and we covered it in attention residue.

End the day with a small deliberate win

The last twenty minutes of your evening have disproportionate weight over how you remember the whole day. Most of us hand those minutes to a feed, which reliably produces a mildly sour ending. Swapping them for one small completed thing, two pages, one short lesson, tidying one surface, changes the remembered day at a very low cost.

Design the finish of any habit you want to keep

Whatever the habit is, ask what its last thirty seconds feel like. Runners who finish with a downhill stretch keep running. People who finish a workout with the hardest set often do not go back. This is also why a streak counter works better than it should: it manufactures a tiny satisfying ending. The mechanics of that are in our piece on the psychology of streaks.

The honest limits

Three caveats, because this is a real effect that gets oversold.

• It shapes memory and choice, not experience. A badly designed hour is still a badly designed hour while you are inside it, and engineering the ending is not permission to make the middle miserable.
• The peak matters as much as the end. A session with a genuinely awful high point will not be rescued by two nice minutes at the finish.
• Effect sizes vary and duration neglect is not absolute. Long experiences do register somewhat, especially very long ones. The claim is that duration counts for much less than intuition suggests, not that it counts for nothing.

Where timing comes in

One useful pairing: peak-end thinking tells you how to shape a session, and time-of-day research tells you where to put it. If your best focus is in the morning, that is where the hard middle belongs, which leaves the easy closing win for a moment when your energy is already fading. We went through the evidence on that in the best time of day to learn.

Why we build lessons this way

Disclosure before I mention our own product: MindSnap is our app, so read this as a design note rather than a neutral recommendation.

Our lessons run about two minutes and they are deliberately built to land, not to trail off. Each one ends on a resolution: the twist in the story, the answer to the question the lesson opened with, the fact you will repeat to someone later. That is a peak-end decision, not a stylistic one. Five flagship collections run as continuing narratives, and unlimited topics sit alongside them, but the rule is the same in all of them, which is that the last line has to be worth arriving at.

It is also why we did not build lessons that stop when a timer runs out. An arbitrary cutoff is a bad ending, and a bad ending is a habit that quietly dies.

Try it on something today. Pick one thing you want to keep doing, and change only the final thirty seconds of it. That is the cheapest behavioural intervention I know.