The room where the dial was
You are lying on your side in a curtained hospital bay in Toronto in the mid-1990s, holding a small box with a dial. The dial moves a marker along a line on a screen, from no pain to extreme pain. Every sixty seconds a prompt asks you one thing: how much does it hurt right now?
That is the first question, and it is asked of the person who is actually in the room. It gets asked again a minute later, and a minute after that, for as long as the procedure lasts — which, across the 154 patients in this study, was anywhere from four minutes to sixty-seven.
Then the procedure ends, and within the hour somebody asks a second question, on a different sheet of paper, on a different scale: the total amount of pain experienced, from "no discomfort" to "awful discomfort". One number. The whole thing.
Two questions. They sound the same, but they are not: the first is answered by the person living the minutes, the second by whoever is left holding the memory. Donald Redelmeier and Daniel Kahneman asked both questions of the same patients and compared the answers.
Before you read what they found, you go first. Below is a dial — nothing here will hurt you, but it is a real episode with a shape, and it will ask you to rate it while it happens. Much later on this page, it will ask you the second question.
Instrument 1 The dial
Skippable, like everything here. If you would rather not sit through it, press "skip the dial" and the page carries on without it. Nothing you do is scored.
Two short episodes, one after the other, each a small, harmless nuisance: a flickering light, a jittery shape, a slow counter. While it runs, keep the dial where it honestly sits: how unpleasant is this right now? Nobody sees it but you.
Your line is drawn as you go. It is yours; it is not a score.
The two lines that started the argument
In the paper — Pain, volume 66, 1996, pages 3 to 8 — there is a figure with two patients on it. Their names are not given. They are called A and B, and between them they have kept psychologists busy for thirty years.
Patient A's colonoscopy lasted 8 minutes; Patient B's lasted 24. Every minute of pain A had, B had too — and then sixteen more. Anyone adding it up would say B had the worse time.
Look at the two lines below before you read another word. Then answer the only question that matters here — not who suffered more, but who remembered it as worse.
Instrument 2 Two patients, one guess
Skippable. Guessing wrong here is the fun part; nothing is recorded as right or wrong about you.
Redrawn by hand from the shape of Figure 1 in Redelmeier & Kahneman (1996). The durations (8 and 24 minutes) and the overall shapes are from the paper; the minute-by-minute values here are an honest approximation, not the original data points.
Which patient rated the total amount of pain as worse, afterwards?
Across the 154 colonoscopy and 133 lithotripsy patients, the remembered rating followed two moments: the worst one, and the last few minutes. How long the procedure lasted made no clear difference. Later research agrees — a 2022 review that pooled 174 results found a large peak–end effect and almost no effect of length.
What predicts how an experience is remembered
Pooled correlation with people's after-the-fact ratings, 174 effect sizes (Alaybek, Dalal et al., 2022)
A longer bar means the measure tracks the remembered rating more closely. The thin line shows the 95% confidence range. The authors call the effect of duration “essentially nil”.
Show the numbers
| Measure | ColonoscAll studies pooled | 95% |
|---|---|---|
| Peak and end | 0.581 | 0.487–0.661 |
| How long it lasted | ≈ 0 | – |
Whether the procedure lasted four minutes or sixty-seven, it was not remembered as worse for being long. This finding has an ugly, perfect name: duration neglect.
Average the worst moment and the last moment, and you get a good guess at how someone will rate an experience afterwards. That shortcut is the peak–end rule. Drag the last minute of a procedure below and watch the two accounts pull apart.
Instrument 3 The last minute
Skippable. This one is a machine, not a mirror — nothing about you is being measured.
Notice what you can do here: you can add three minutes of genuine, real discomfort to somebody's afternoon and make them remember the afternoon as better. Hold on to that thought. In act seven, a hospital does exactly that, on purpose, to 682 people.
Thirty-two students and a bucket of cold water
Three years before the hospital paper, in 1993, Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier published a study with a title that reads like a dare: When More Pain Is Preferred to Less: Adding a Better End. Thirty-two male University of California students, aged 19 to 39, median age 22 and a half.
Each did two trials. In the short one, a hand went into 14°C water — cold enough to ache — for 60 seconds. In the long one, the other hand did the same 60 seconds and then stayed 30 seconds more while the water warmed slowly to about 15°C: still painful, but less so.
So the long trial contains the entire short trial, plus half a minute of extra cold on top. By any honest accounting it is strictly worse. Then the experimenter asks the question you have already answered on this page:
"Suppose we paid you to come back tomorrow to repeat just one of the two cold-water trials that you've experienced today. Which one would you choose?"Kahneman, Fredrickson, Schreiber & Redelmeier, Psychological Science 4(6), 1993, pp. 401–405
What they chose
Twenty-two of the thirty-two — 69% — chose to repeat the long trial, with more cold and more ache. They had not forgotten which was longer. They chose the better memory, because the last thing that happened was the water getting kinder.
This is the moment the whole thing stops being about hospitals. If a warmer half-minute at the end can flip a preference, then the story you carry around about your own week, your own job, your own summer is not a recording. It is a summary — and summaries are written by editors with strong opinions and very little patience.
Speed round Memory's verdict
Skippable. A quick game about made-up patients; nothing about you is measured.
Two episodes appear side by side. Tap the one the patient will remember as worse before the timer runs out. The total pain is shown too — it is a trap.
Now the second question — the one asked afterwards
A while ago you sat through two small episodes with a dial in your hand. The dial is finished with you. What follows is the other question, the one the hospital asked within the hour, on a different sheet of paper: not how it felt at any moment, but the total amount.
Answer it from memory. Do not scroll back up to look at your line. That would be cheating — and more to the point, it would be answering with the wrong self.
Instrument 4 The second question
Skippable. There is no right answer, and the number you give is not compared to anyone else's.
Now try a bigger version, with something that matters to you. Choose an episode of your own life — a day, a school year, a job, a trip, a winter. Draw how it felt as it went, not how it turned out.
Instrument 5 Your own line
Skippable, and deliberately unfussy: pick something ordinary if you'd rather. A commute, a bad film, a good summer. Nothing here is stored anywhere but in this page, and nothing about it is scored.
Drag across the chart to draw. Left is the beginning, right is the end. High is good, low is bad — this line runs the friendly way up, unlike the hospital's.
Six hundred million moments, and the handful that got kept
In February 2010, Kahneman stood on a TED stage and put it in the plainest words he ever used for it. There is an experiencing self, who lives in the present and knows the present. And then, he said, there is a remembering self — and the remembering self is the one that keeps score, and maintains the story of our life.
Then he did the arithmetic. The psychological present lasts about three seconds, so a month holds about 600,000 moments and a life about 600 million. Only a tiny handful survive as anything you could name.
"We actually don't choose between experiences, we choose between memories of experiences. And even when we think about the future, we don't think of our future normally as experiences. We think of our future as anticipated memories."Daniel Kahneman, TED, February 2010
He told a story about a man listening to a glorious symphony recording that ended with a dreadful screech. The man said the screech had ruined the whole experience. Kahneman's reply: it ruined the memory — the twenty minutes of music were real, but the memory was all he got to keep.
Below is a month of your life, drawn at the correct scale: 600,000 dots, one for each three-second moment, most of which you have already lost forever and were never going to keep. Find the few that stayed. You will not need many rows.
Instrument 6 A month, at actual size
Skippable. Write as few as you like — one is a perfectly good answer, and so is none. Please keep it to things you don't mind seeing on a screen.
600,000 dots. Roughly the last thirty days, at three seconds each. Click anywhere to light one up and give it a name.
Whatever number you stopped at, the dots you did not light are not wasted. You lived every one of them. That is what the experiencing self is for — and it is the only one of the two of you that ever actually gets to be there.
The James Dean effect
In 2001, Ed Diener, Derrick Wirtz and Shigehiro Oishi took the same instrument and pointed it at something much bigger than a colonoscopy: an entire life. They described a life to people and asked how desirable it was. Then they described the same life to other people, with some extra years added on the end — years that were pleasant, just less wonderful than what came before.
Rate the two lives yourself, before the result. The second one contains everything in the first one, and then some more good years. Not tragic years. Good ones.
Instrument 7 Two lives, one dial each
Skippable. This one is about a stranger, not about you — though the answer tends to say something anyway.
Life one
Someone lives a wonderful life. Close friends, work they love, a great deal of laughing. It is, by any measure, a marvellous run. Then it ends, suddenly and painlessly, at its height.
Life two
Exactly the same wonderful life, the same friends, the same laughing — and then, after it, some additional years that are mildly pleasant. Not bad years. Not sad ones. Just ordinary: quieter mornings, smaller days. Then it ends.
Diener and his colleagues found this in three studies, plus its mirror image: a terrible life with some extra moderately bad years was rated better than the same life ending abruptly — the “Solzhenitsyn effect”. When the extra years went in the middle instead of at the end, the effect was much weaker. The ending gets the vote.
Which is strange, and a bit beautiful, and worth saying out loud: nobody in these studies is adding up the years. They are reading the shape.
What a hospital did with this, and the uncomfortable part
Between November 1994 and December 1995, Donald Redelmeier, Joel Katz and Daniel Kahneman ran the experiment this finding demands, with 682 colonoscopy patients in Toronto. 345 got the usual procedure; 337 got one addition: at the end, the instrument tip stayed in place for up to three extra minutes — no suction, no inflation, no extra medicine. Mildly uncomfortable, but much less so than what came before.
Both groups spent about the same time overall (27.6 against 26.8 minutes), but the gentler-ending group rated the whole thing as less painful: 4.4 against 4.9 (p = 0.006). Over a median of 5.3 years, 53% of them came back for their next colonoscopy, against 48% of the others. On its own that gap could be chance; after adjusting for other factors, their odds of returning were 1.41 times the others'.
"Last impressions may be lasting impressions."Redelmeier, Katz & Kahneman, Pain 104 (2003) 187–194
Their practical suggestion was that doctors might use special gentleness and care at the end of an aversive procedure. Their own paper calls the result modest and calls itself a proof of concept.
Now the uncomfortable part. To improve the memory, they added real discomfort to real people's afternoons: the experiencing self paid three minutes so the remembering self could have a better story — and a better story brings people back for a test that catches cancer early. You can think that trade was worth it and still feel how big a lever it is.
Which is why the last instrument on this page is a small one, and points at your own day rather than anybody else's.
Instrument 8 The ending you can choose
Skippable. Keep it small and ordinary — small and ordinary is the whole point.
Not to trick yourself. The experiencing self still has to live through the middle, and no ending fixes a genuinely bad hour. But endings are cheap, and you own yours.
Six honest wobbles
Every edition of this page carries a list like this, and today's is long, because the peak–end rule is the sort of finding that gets repeated at conferences long after the arguments about it have started.
1. Thirty-two men, one university
The cold-water study had 32 male students from one American university. The colonoscopy study had 154 patients at one Toronto hospital in one decade. These are the foundation stones of a claim about human memory, and they are not large.
2. A whole day does not obey it
Tamara Miron-Shatz (Emotion, 2009) asked 810, 820 and 805 women in the United States, France and Denmark to rebuild the previous day episode by episode. Their rating of the whole day was best predicted by how they felt, weighted by how long each feeling lasted — exactly the sum that duration neglect says we skip. The final episode added essentially nothing.
3. Holidays don't obey it either
Kemp, Burt and Furneaux (Memory & Cognition, 2008) sent 49 students on holidays averaging seven days and collected a daily text message about how happy they had been. The peak–end rule "was not an outstandingly good predictor" of how they later rated the trip; the happiness of the most memorable day did better. Duration, though, still had no effect at all.
4. Duration comes back the moment you have to choose
Dan Ariely and George Loewenstein (JEP: General, 2000) showed that duration is not simply invisible. How much weight it gets depends on what you are asked to do. Put two experiences side by side, or make people choose or pay, and duration matters a great deal more than it does in the classic task of rating one thing on its own — which happens to be the task that produced duration neglect in the first place.
5. But the meta-analysis backs it
Fairness cuts both ways. Alaybek, Dalal and colleagues pooled 174 effect sizes (Organizational Behavior and Human Decision Processes, 2022) and found strong support: a large peak–end effect (r = 0.581) that held across many conditions, while the effect of duration was “essentially nil”.
6. And this page is not a measurement of you
Two dial traces and a hand-drawn line are an illustration of a method, not a reading of your mind. With one episode and one number there is no way to tell a pattern from an accident — that is not a flaw in you, it is arithmetic. Nothing here diagnoses anything, scores anything, or knows anything about you that you did not just type in.
A short history of the two selves
What you said, back in your own words
This panel does not know anything about you. It can only read what you typed and moved on this page, and say it back. Then it names one pattern — and you get to throw it out. A mirror is not a verdict; it is just a thing you look into and disagree with.
Eight questions — on the story, not on you
Everything above is a mirror: unscored, unjudged, yours. What follows is a quiz with right answers, and it is only about other people's experiments, dates and numbers. Nothing you wrote about yourself is graded.
One last question, unscored
Kahneman used to ask people a thought experiment, and it is the best thing to leave you with. Imagine that at the end of your next holiday, every photograph will be destroyed, and you will be given a drug so that you will not remember any of it. Would you still choose the same holiday?
Most people find they would choose differently — which means most of what we plan, we are planning for the one of us who will not be there.
So: what would you do tomorrow, if nobody — including you — were ever going to remember it?