In a classic experiment, people preferred to repeat 90 seconds of painfully cold water rather than 60 seconds of the same pain because the longer trial ended just one degree warmer, revealing that memory often judges an experience by its worst moment and its ending rather than by how long it lasted

There is the pain a person lives through, and then there is the pain their memory files away for later. The unsettling part of the classic cold-water experiment is that those two versions of an experience do not always agree.

In the study, published in Psychological Science in 1993 by Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier, participants put one hand in painfully cold water. In one trial, the water stayed at 14C for 60 seconds. In another, it stayed at the same painful temperature for 60 seconds, then continued for another 30 seconds while the temperature rose slightly to 15C.

On paper, the second trial was objectively worse. It contained all the discomfort of the first trial, plus half a minute more. Nothing was removed. The person simply got a slightly less painful ending.

Yet when participants were later asked which trial they would rather repeat, many chose the longer one. Their memory had not added up the total amount of suffering. It had weighted the story differently: the worst moment mattered, and the ending mattered. Duration mattered less than it should have if the mind were keeping a faithful running total.

That is why the experiment has become one of the clearest demonstrations of what later became known as the peak-end rule. We do not remember many experiences as continuous records. We compress them into highlights, low points and endings. Then we use those compressed memories to decide what we want to avoid, repeat, recommend or forgive.

The remembering self is not a stopwatch

The cold-water result is often described as irrational, but that word can flatten what is actually happening. The brain is not a measuring instrument. It is a decision system trying to turn a flood of moments into something usable.

If every uncomfortable meeting, medical procedure, product demo, customer support call or difficult conversation had to be stored with perfect second-by-second fidelity, memory would be unusably detailed. Instead, the mind simplifies. It asks: how bad did this get? How did it end? What story should I carry forward?

A separate paper by Fredrickson and Kahneman on duration neglect in retrospective evaluations helped formalise the same point: the remembered value of an episode can diverge sharply from the sum of its moment-by-moment feelings. A longer unpleasant experience may be remembered more kindly than a shorter one if it ends better. A mostly pleasant experience may be spoiled in memory by one sharp low point at the end.

That distinction matters because people rarely make choices from the version of life they actually experienced in real time. They make them from the version they remember.

A customer may spend 40 minutes getting useful help from a company and still describe the whole interaction as bad if the last two minutes feel dismissive. An employee may remember a demanding project with surprising warmth if the final stretch included recognition, control and a clean landing. A candidate may walk away from an interview less influenced by the average tone of the hour than by the one question that felt humiliating and the way the interviewer closed the call.

The total duration is not irrelevant. But memory does not treat duration like an accountant. It treats experience more like an editor.

Why endings become so powerful

Endings have a special status because they are where an experience becomes a memory. They tell the brain what kind of episode it has just survived.

This is not just a laboratory curiosity. Redelmeier, Katz and Kahneman later tested the logic in a clinical setting in a randomized colonoscopy trial. Some patients had the procedure extended briefly at the end in a way that was designed to be less uncomfortable than the peak of the procedure. The intervention increased the total duration, yet it improved remembered discomfort for some patients and affected later willingness to return.

That finding should be handled carefully. It is not a general instruction to prolong unpleasant experiences, and it does not mean pain is somehow cancelled by a better finish. But it shows how consequential the final stretch of an experience can be when people later decide what happened to them.

In business and technology, this is a quiet design problem. Products and workplaces often focus on throughput: shorten the process, reduce the number of clicks, make the call faster, move the meeting along. Those things matter. But a faster experience that ends in confusion, indifference or unresolved tension may be remembered worse than a slightly longer experience that ends with clarity.

Anyone who has left a well-run support chat knows the feeling. The problem may have been irritating. The wait may have been longer than ideal. But if the final message is specific, calm and useful, the story the customer carries away changes shape. The end does not erase the difficulty. It gives it a different meaning.

The workplace version of the cold-water tank

The peak-end rule is especially useful because so much modern work is made of episodes. A product launch, a performance review, a quarterly planning cycle, a sales call, an onboarding week, a layoff conversation, a failed experiment: each becomes a remembered unit.

People do not later recall these episodes by averaging every minute. They remember the moments of highest stress, the moment they felt most seen or exposed, and the way the episode ended.

This helps explain why some teams carry scars from events that leaders think were handled “mostly fine.” From the leader’s view, the difficult moment lasted ten minutes inside a much longer process. From the employee’s memory, that moment may be the process. If the ending then feels rushed, defensive or bureaucratic, the whole episode can harden into distrust.

The reverse can also happen. A hard project can become a source of pride if its ending gives people a chance to make sense of it. A candid retrospective, real appreciation, and a clear decision about what changes next can alter the remembered shape of the work. The suffering does not disappear. But it is no longer pointless suffering.

This is not sentimentality. It is design. The final five minutes of a meeting, the follow-up after a complaint, the way a manager closes a hard conversation, the last screen of a signup flow, the goodbye email after someone leaves a company: these are memory-forming moments.

What the research does and does not say

The peak-end rule is powerful, but it is not magic. A 2022 meta-analysis in Organizational Behavior and Human Decision Processes found support for peak and end effects while also making clear that the pattern is not identical across every kind of experience, measure or setting. Sometimes average intensity matters. Sometimes duration matters. Sometimes the most recent moment matters because it is vivid, not because of a universal law.

That nuance is important. The cold-water experiment should not be turned into a cheap management trick. It does not say people can be manipulated into liking bad treatment if the ending is polished. It says memory is structured, and that structure deserves respect.

If an experience has a genuinely awful peak, fix the peak. If a process takes too long, shorten it. If the end is chaotic, give it a better landing. The lesson is not to decorate discomfort. It is to understand that people leave an experience through a door, and the state of that door becomes part of the memory.

For founders, managers and product teams, the practical question is simple: when someone reaches the end of an interaction with us, what are they being asked to remember?

If the answer is confusion, silence, embarrassment or friction, the episode may be stored that way even if most of it was competent. If the answer is clarity, dignity and resolution, the same episode may be carried forward differently.

The 90-second cold-water trial was not better than the 60-second one. It hurt for longer. But it ended slightly less badly, and that was enough for memory to prefer it. The finding is uncomfortable because it exposes a gap we live with every day: experience happens in time, but memory turns it into a story.

And very often, the story is what decides what we choose next.