Peak-End Rule: How Endings Distort Your Memory Of Experiences

The 1996 study by Redelmeier, Katz, and Kahneman is worth examining in detail because the design is so clean and the finding is so counterintuitive.

Colonoscopy patients were randomly assigned to one of two conditions. In the standard procedure (control group), the colonoscopy ended when the clinician completed the procedure and removed the scope. In the experimental condition, after completing the procedure, the clinician left the scope in place for approximately 60 extra seconds with minimal movement. This is not painful in itself but is mildly uncomfortable — and it extends the total duration of the procedure.

At regular intervals during the procedure, patients rated their current discomfort on a scale. After completion, they rated the global experience. One year later, the researchers measured rates of compliance with recommended follow-up procedures.

Results: Patients in the extended condition reported lower global discomfort — despite more total discomfort minutes — and were more likely to comply with follow-up recommendations at the one-year mark.

The mechanism: the extended group's final moments were less intense than the peak moments of the active procedure. The ending was "softer." Because memory weights the peak and end heavily, this softer ending produced a less aversive remembered experience, which then produced better health behavior downstream.

The experienced discomfort was greater. The remembered discomfort was less. And it was the remembered version that drove future decisions.

Kahneman built a framework around this finding: the distinction between the experiencing self and the remembering self. This is not a metaphor — it's a genuine cognitive distinction with measurable behavioral consequences.

The experiencing self is the one living in the present moment. It has interests: it prefers less pain to more, shorter bad experiences to longer ones, more pleasure to less. Duration matters to it. Every moment counts.

The remembering self is the one that builds narrative summaries of experiences and uses those summaries to make choices. It has different interests: it cares about peaks and endings. Duration barely registers — a phenomenon Kahneman called "duration neglect." Ten minutes of bad pain followed by five minutes of mild discomfort feels worse in memory than twenty minutes of bad pain followed by five minutes of mild discomfort, because the ending is the same and the peak is what it is.

Most of life as we live it is an experiencing-self phenomenon. But most of what drives our choices is a remembering-self phenomenon. We decide to return to a restaurant, take a vacation again, stay in a relationship, leave a job — based on remembered experiences. Those remembered experiences are not accurate averages of the actual experiences. They are peak-and-end reconstructions.

This creates systematic distortions in how we make decisions. We over-optimize for endings. We under-weigh sustained mediocrity (long stretches of mildly unpleasant experience get discounted in memory because they never hit a dramatic peak). We over-react to bad endings of otherwise good experiences.

Duration neglect is the most underappreciated component of the peak-end rule. In cold pressor experiments (where participants submerge their hand in cold water), participants reliably prefer a longer trial with a less intense final phase over a shorter trial with the same intensity throughout. More total pain, but a better ending, produces a better remembered experience.

When this is applied to temporal well-being research, the implications are troubling. Kahneman ran the Day Reconstruction Method research — asking people to track their emotional experience through the day — and found significant gaps between experienced well-being (how people feel moment-to-moment, averaged across time) and remembered well-being (how people evaluate how their life is going).

People with high incomes often report high life satisfaction (a remembering-self evaluation) without showing proportionally higher experienced positive affect. People in certain situations report low life satisfaction without having commensurately worse daily emotional experience. The two measures are correlated but far from identical, and they're produced by different cognitive processes.

The question "are you happy?" asks the remembering self. The answer reflects narrative, peak moments, how things ended, and comparison to expectations — not the actual texture of daily experience. Neither measure is wrong, but they measure different things.

The peak-end rule has been discovered and applied by industries that design experience at scale, often without naming it.

Hospitality: Hotels and airlines invest disproportionately in the arrival and departure experience. A warm, smooth check-in and a gracious checkout produce better reviews than a mediocre stay with extraordinary midpoint service, even when total service quality is similar. The endpoints anchor the memory.

Theme parks: Disney's design philosophy has long incorporated something like the peak-end rule — the park is structured so that the most spectacular moments are distributed throughout and the exit experience (merchandise, photos, final impressions) is carefully managed. You leave through Main Street, USA, which is designed to be warm and cheerful.

Healthcare: Some health systems have begun redesigning discharge processes based on this research. The last minutes of a hospital visit — often logistically chaotic, depersonalized, rushed — leave a disproportionately strong impression. Improving the discharge experience can shift patient satisfaction scores meaningfully even when nothing else changes.

Retail: "Surprise and delight" moments late in a customer journey — a handwritten note in a package, an unexpected upgrade at final checkout — exploit the peak-end rule deliberately. The peak can be manufactured.

The darker applications: subscription services that make cancellation difficult create an artificially bad ending to the relationship, which ironically can produce better word-of-mouth than easy cancellation (because a hostile exit is memorable, and not always in ways that hurt the brand as much as you'd expect). Difficult endings that include any positive resolution can paradoxically produce stronger loyalty than smooth experiences. Casino design deliberately creates win peaks — the slot machine that occasionally produces a significant payout — knowing those peaks will dominate the memory of an evening where the overall outcome was losing.

The most consequential application of the peak-end rule is in how you evaluate your own autobiographical experience.

Your memory of relationships, jobs, periods of life, places you've lived, educational experiences — all of it is being filtered through peak-and-end processing. This creates specific predictable distortions:

Bad endings contaminate good experiences. A relationship that was largely good but ended badly will often be encoded as a bad relationship. The ending's weight is disproportionate to its duration. People write off years of genuine positive experience because the ending was painful — which may be accurate (some experiences are defined by how they end) or may be a memory artifact.

Peak moments define periods. A mediocre year with one extraordinary moment will often be remembered as a good year. A good year with one awful experience may be remembered as worse than it was. When you assess whether "the startup years" or "the time I lived in X" were good, you're running peak-and-end processing without knowing it.

Duration discount. Extended periods of sustained moderate experience — good or bad — are systematically compressed in memory. Years of steady, satisfying but unspectacular work are remembered as less than they were. Years of low-grade unhappiness are also compressed, which can make past situations feel retrospectively better than they were.

The practical corrective: keep records. Journals are not just for expression — they're for accuracy. If you have a contemporaneous account of what experience was actually like, you can compare it to your remembered version. The gaps are informative.

Also: be deliberate about endings. If you know endings carry disproportionate weight in memory, you can engineer them. End projects with a clear celebration. End working sessions before you're depleted. End visits on a high note. This is not manipulation — it's designing the memory that will inform your future choices.

Kahneman himself raised the uncomfortable question: whose interests should you optimize for?

If you optimize for the remembering self — engineering peaks and endings — you may be producing better memories at the cost of actual experience. The colonoscopy patients who had a better memory had a worse experience. Extending a painful procedure to soften the ending serves the remembering self's preferences while costing the experiencing self additional minutes of discomfort.

This is not hypothetical. Every time you "tough out" an unpleasant experience because the peak or ending will be worth it, you're making a trade. Sometimes that trade is right — the experiencing self endures something the remembering self will value. Sometimes it isn't — you sacrifice real experience quality for a memory that will be reconstructed anyway, possibly inaccurately.

There's no clean resolution here. The key is awareness: knowing that you have two interests in any experience, they're not always aligned, and the remembering self is the one making your future choices. Use that awareness to be deliberate rather than just running the memory architecture blind.

The experiencing self is living your life. The remembering self is narrating it. Both matter. Most people have only met one of them.