The Peak-End Rule
Here is the uncomfortable thing the Peak-End Rule forces you to confront: the version of you that had an experience and the version that remembers it are two different witnesses, and they do not agree. The one who lived it counted every minute. The one who remembers it threw most of those minutes away and kept two — the worst (or best) instant, and however it ended. Those two snapshots become "what happened," and from then on, that's the story you act on. You don't make decisions based on the life you lived. You make them based on the highlight reel your memory edited, badly, while you weren't looking. This is one of the most disorienting findings in cognitive science precisely because it's so easy to verify in your own life once you know what to look for. Let's pull it all the way down — and let's be honest about which parts are bedrock and which parts are softer than the popular retellings admit.
What it is (and where it came from)
The Peak-End Rule is the claim that when people retrospectively judge how good or bad an experience was, the judgment is dominated by two moments: the peak (the most intense point, good or bad) and the end (how it felt as it wrapped up). The average of all the moments, and crucially the total duration, get largely ignored.
That last part — duration getting ignored — has its own name: duration neglect. If a painful experience lasts ten minutes or twenty, your remembered sense of how bad it was barely changes, as long as the peak and the ending are comparable. Time, the thing you actually spend, mostly drops out of the remembered verdict.
The work belongs primarily to Daniel Kahneman — the Israeli-American psychologist who won the 2002 Nobel Prize in Economics for the broader research program he built with Amos Tversky — together with several collaborators in the early-to-mid 1990s. The most-cited foundational paper is Kahneman, Fredrickson, Schreiber, and Redelmeier (1993), "When More Pain Is Preferred to Less: Adding a Better End," published in Psychological Science. The phrase and the framework get sharpened across Kahneman's subsequent work, and he gives the most accessible full treatment in his 2011 book Thinking, Fast and Slow, where he frames it through what he calls the experiencing self versus the remembering self. That distinction is the philosophical heart of the whole thing, and we'll come back to it, because it's bigger than a memory quirk.
The deeper conceptual frame Kahneman built around this is the split between experienced utility and remembered utility — borrowing "utility" from economics. Experienced utility is the real-time, moment-by-moment quality of an experience: how good or bad it actually feels as it unfolds. Remembered utility is what you think it was, after the fact, summarized. Classical economics quietly assumed these were the same thing — that the value of an experience is roughly the integral of its moments, summed over time. Kahneman's program showed they come apart, systematically, in a direction you can predict. And here's the sting: we make our future choices based on remembered utility, not experienced utility. The remembering self is the one holding the pen when you decide whether to do something again.
The real science / evidence
There are two studies that do almost all the heavy lifting here, and you should know them by heart because they're clean, they're real, and they're the ones everything else leans on.
The colonoscopy study — Redelmeier & Kahneman (1996). Published in Pain, this is the one that turned a lab curiosity into something nobody could dismiss. In the era before routine sedation made the procedure painless, Donald Redelmeier (a physician and researcher in Toronto) and Kahneman had real patients undergoing colonoscopies report their pain in real time, every 60 seconds, on a scale, throughout the procedure. So for each patient they had a full moment-by-moment record — the actual experienced pain, minute by minute. Afterward, patients rated the overall badness of the whole experience and their willingness to return.
The result is the heart of the entire field. The patients' retrospective judgments of how bad the colonoscopy was correlated strongly with the average of the peak moment and the final moment of pain — and correlated poorly with the total duration. A procedure that lasted four minutes and one that lasted over twenty could be remembered as roughly equally bad, if their peak-and-end profiles were similar. Two patients are often contrasted: one had a short procedure that ended at a high level of pain; another had a much longer procedure — more total suffering by any honest accounting — but it tapered off to a mild ending. The second patient, who objectively endured more total pain, remembered the experience as less bad. More pain, better memory. The title of the 1993 paper says it outright: "When More Pain Is Preferred to Less."
There's a second, even more provocative layer that Kahneman ran (the 1993 cold-pressor experiment, below, is the controlled version, but the colonoscopy program inspired a clinical follow-up): the logical implication is that deliberately extending a painful procedure with a few extra minutes of milder discomfort at the end would leave patients with a less awful memory — and more willing to come back for the screening that might save their life. That's a genuinely strange ethical knot: you could add real suffering to a person's day and improve their life, because the added mild ending rewrites the memory and keeps them coming back for preventive care. This was actually tested: Redelmeier, Katz and Kahneman (2003, Pain) randomized 682 colonoscopy patients, and those given a short, milder extension at the end remembered the procedure as less unpleasant; the increase in returning for a repeat colonoscopy was small and not statistically significant. So I want to be careful here: the strong "we should lengthen procedures" framing is not established clinical protocol, and you should treat it that way. What's solid is the measurement: remembered badness tracked peak-and-end, not duration.
The cold-pressor study — Kahneman, Fredrickson, Schreiber & Redelmeier (1993). This is the controlled, within-subjects version, and it's the cleaner experiment because the same people did both conditions. Participants immersed a hand in painfully cold water. In the short trial, they held it in 14°C water for 60 seconds. In the long trial, they held it in 14°C water for 60 seconds and then kept it in for 30 additional seconds while the water was slightly warmed — to about 15°C — still uncomfortable, just marginally less so. So the long trial contained all the pain of the short trial plus 30 extra seconds of (lesser but still real) pain. By any rational accounting, the long trial is strictly worse — it's the short trial plus more suffering.
Then participants were told they'd have to repeat one of the two trials, and asked which they'd choose. A substantial majority — roughly two-thirds in the original report — chose to repeat the long trial, the one with objectively more total pain. Why? Because it ended better. The slightly-warmer tail gave it a gentler ending, and the remembering self, weighing peak-and-end, judged it less aversive. People volunteered for more pain to get a better memory of it. That's the rule in its purest experimental form, and because it's within-subjects (same people, both conditions, choosing freely), it's hard to wave away.
Beyond these two anchors, peak-end and duration-neglect effects have been reported across a range of paradigms — aversive film clips, unpleasant sounds, vacations, and consumer experiences — which is why the rule entered the textbooks. But I want to flag the honest caveat early, because the rest of this document depends on you holding it: the foundational pain studies are robust and replicated in their core finding, but the breadth of the rule — how far it generalizes to long, rich, multi-day experiences and to pleasant rather than painful ones — is genuinely more contested than the popular business-book version lets on. We'll get to the debates. For now: the anchor studies are real and strong. The empire built on top of them is shakier in places.
The mechanism
Why would a mind be built this way? It looks like a bug. It's better understood as a feature with a steep cost.
The remembering self isn't trying to give you an accurate accounting ledger of your life. It's trying to do something cheaper and more useful for survival: compress an experience into a small, storable, decision-ready summary. You cannot carry a minute-by-minute recording of every experience you've ever had — the storage and retrieval costs would be ruinous. So the brain does what any good compression algorithm does: it keeps the most informative samples and discards the redundant middle.
For learning whether to approach or avoid something in the future, two data points are disproportionately informative. The peak tells you the maximum intensity this category of experience can deliver — the worst it can hurt you, the best it can reward you. The end tells you the trajectory — where things were heading, which for a predictive organism is often more useful than where they were on average, because the recent state is the best guess about the next state. Duration, by contrast, is comparatively uninformative for the avoid/approach decision. Knowing that something was painful for nineteen minutes rather than four doesn't change the basic verdict "this hurts, and it was getting better/worse at the end" very much. So evolution, or just the economics of memory, throws duration away.
There's a complementary mechanism worth naming. Real-time pain (or pleasure) is processed by the experiencing self — moment to moment, in the present. But memory encoding is discrete and event-bounded: we store episodes with salient markers, not continuous footage. The peak is salient by definition (maximum intensity = maximum encoding strength), and the end gets privileged encoding partly through recency — it's the last thing in working memory when the experience closes and the summary gets written. The "summary verdict" is essentially constructed at the boundary of the episode, weighting whatever is most available at that moment. Duration has no salient marker. Nothing about "this lasted a long time" lights up the encoding machinery the way a peak or a final state does.
So the rule isn't a malfunction. It's the predictable output of a system optimized for cheap, decision-relevant storage rather than faithful accounting. The cost is that the metric your memory optimizes for is not the metric your life is actually lived in. You live in experienced utility — every minute, summed. You decide in remembered utility — peak and end. The gap between those two is where a lot of human misery and a lot of bad decisions live.
Sharp examples
The clinical studies are the proof. These are where you feel it.
The vacation that "wasn't worth it." Two weeks of genuinely lovely days — but the flight home was a nightmare of delays, a missed connection, and a screaming argument at the rental car counter. Ask a month later how the trip was, and a lot of people will summarize it as a disaster, or at least "ruined at the end." The experiencing self had thirteen good days. The remembering self kept the peak and the wreckage of the end, and the wreckage was recent. The thirteen good days were real — they happened, the experiencing self enjoyed every one — but they barely survive into the verdict. Duration neglect, live.
Why you'd go back to a worse vacation. Kahneman's own famous thought experiment: imagine a wonderful trip, and then imagine that at the end all your photos are destroyed and you're given a drug that erases the memory entirely. Many people report the trip would feel almost pointless — they'd be far less willing to take it. But notice what that reveals: you're not valuing the experience, you're valuing the memory of it. The experiencing self had exactly the same beautiful time either way. The fact that the memory-erasure makes it feel worthless tells you which self is actually running your choices.
The last five minutes of the massage / the dentist / the meeting. A pleasant massage that ends with an abrupt, jarring "okay, time's up" can be remembered as worse than a shorter one that ends with a gentle, slowing close. Skilled service providers know this in their bones, even if they've never heard the term: end well, and the whole thing is remembered well.
The relationship that "was always bad." This one cuts deepest. A relationship ends in a brutal, drawn-out, bitter breakup — and afterward the person reconstructs the entire history as having been bad all along. But it wasn't. There were genuinely happy years. The experiencing self lived them. The remembering self, weighting the agonizing peak and the bitter end, overwrites the average and rewrites history. People do this to careers, to cities they lived in, to friendships. The end becomes the lens through which the whole thing is re-seen — and the re-seeing feels like truth, not distortion. That is the dangerous part: the edit doesn't feel like an edit. It feels like clarity.
Why finishing strong is not just motivational fluff. The end of any designed experience — a course, a customer onboarding, a performance, a meal — carries weight wildly out of proportion to its duration. The closing two minutes can do more to set the remembered verdict than the preceding hour. This is real leverage, and it's the one genuinely actionable thing the rule hands you. But it's also exactly where it gets oversold, so hold on.
Where it gets distorted / oversold (the honest cut)
This is the section the LinkedIn posts skip. If you only take the triumphant version, you'll over-trust it and get burned.
Distortion 1: "Just nail the ending and nothing else matters." No. The rule says peak and end, and the peak still has to be earned by the substance of the experience. You cannot have a mediocre product, slap a delightful checkout animation on it, and ride the Peak-End Rule to glory. A great ending on a genuinely bad experience produces a less bad memory, not a good one. The colonoscopy patients with better endings still remembered the procedure as unpleasant — just less so. The rule modulates the verdict; it doesn't manufacture one from nothing.
Distortion 2: Treating it as a universal law instead of a tendency. Peak-end is a robust statistical tendency under certain conditions, not an iron law that overrides everything. Duration neglect is strongest for relatively short, single-episode experiences of fairly continuous affect — the kind you can hold in one mental "chunk." For long, complex, multi-phase experiences (a five-year job, a marriage, raising a kid), people demonstrably do integrate more than two moments, and other heuristics kick in. The rule weakens as the experience gets longer and more narratively complex — which is ironic, because that's exactly where business writers most love to invoke it.
Distortion 3: Ignoring the replication and generalizability questions. This needs to be said plainly, because cognitive science is in the middle of a serious replication crisis and intellectual honesty demands you know where the soft ground is. The original pain studies (colonoscopy, cold-pressor) hold up well — they've been examined and the core peak-end/duration-neglect effect in those aversive paradigms is considered solid. But the generalization to pleasant experiences, to remembered enjoyment of varied real-world events, and to the predicted choices people make based on those memories is more mixed. Some attempted extensions find the effect attenuated, swamped by other factors, or absent. The rule is real; "the rule explains everything and replicates everywhere" is not established and should be treated with suspicion. When someone cites peak-end to justify a sweeping claim about a complex year-long experience, your honest answer is: the foundational pain work supports the mechanism, but this particular extrapolation is beyond what the evidence cleanly shows.
Distortion 4: The manipulation read. Service-design and marketing culture has absorbed peak-end as a tool for engineering memories — and there's a legitimate version (genuinely improving how experiences close) and a cynical version (papering over a bad product with a manufactured peak and a sweet ending to inflate reviews). The cynical version exploits the exact gap the rule describes: it optimizes the remembered utility while leaving the experienced utility — the part people actually live through — untouched or worse. That's not service design; that's memory fraud. Knowing the rule means you can also defend against it: when something ended suspiciously well, ask the experiencing-self question — minute by minute, how was this actually?
The debates & frontier
Is it really the end, or the trend? A live question is whether the "end" effect is about the final moment or about the direction of change leading into the end — improving versus deteriorating. The cold-pressor result is consistent with both readings (the end was both lower-pain and on an improving trajectory). Some researchers argue the real driver is the slope — that we're built to care intensely about whether things are getting better or worse, and "the end" is just where that slope is most recently sampled. If that's right, the deeper principle isn't "endings matter" but "trajectories matter," and the end is privileged only because it's the freshest reading of the trajectory.
How does narrative interact with it? Humans don't store experiences as raw affect curves; we store stories. There's an argument that peak-end is downstream of a more fundamental tendency to remember experiences as narratives with a meaningful resolution — and that a satisfying ending in the story sense (not just lower pain) is what gets weighted. This blurs the line between a memory heuristic and a narrative one.
The experiencing-self problem and the measurement of well-being. Kahneman spent his later career wrestling with the policy implication: if the remembering self and experiencing self disagree, which one should we try to make happy? Should public policy and personal choice optimize the life you actually live moment-to-moment, or the life you remember and report? He developed the Day Reconstruction Method partly to get at experienced well-being, because standard "how satisfied are you with your life" surveys query the remembering self and inherit all its distortions. This is unresolved and genuinely deep: there may be no fact of the matter about which self "you" are, and the two can want different lives. Late in his life Kahneman reportedly leaned toward the view that for practical purposes the remembering self is the one that governs our choices and therefore the one that effectively rules — a somewhat bleak concession, given that it's the less accurate of the two.
Boundary conditions are the real frontier. The honest cutting edge isn't "does peak-end exist" — it does — but "under exactly what conditions, and how strongly, and overridden by what?" That's the unglamorous, important work, and it's where the replication-era scrutiny is rightly focused.
How to apply it
The rule earns its keep in two opposite directions: designing experiences for others, and defending your own judgment against the distortion. Most write-ups only cover the first.
- Engineer the peak and the end — honestly. If you build experiences (a service, a course, an event, a client relationship, a meal, a date), invest disproportionately in one genuine high point and a strong, deliberate close. A weak, abrupt, or sour ending will be remembered out of all proportion to its length. End on the upswing. This is real leverage as long as the substance underneath is real — peak-end polishes a good experience; it cannot fake one.
- Never end on the low. Whatever it is — a workout, a workday, a hard conversation, a study session — the last few minutes get over-weighted in the memory and therefore in your future willingness to do it again. End a workout with a set you can win. End a hard conversation with one sincere moment of repair. You're not being dishonest; you're refusing to let an arbitrary final moment author the whole verdict.
- Interrogate your own retrospective verdicts. When you find yourself concluding "that whole trip was terrible," "that job was always miserable," "that relationship was a mistake from day one" — stop and run the experiencing-self check. Minute by minute, day by day, how was it actually, before the end rewrote it? Often the verdict is a peak-and-end artifact, and the lived reality was far more mixed. This single habit defends you against rewriting your own history in the color of how things happened to end.
- Distrust decisions driven purely by how something ended. A spectacular finish (or a bitter one) is a known memory-inflator. Before acting on "I'd never do that again" or "that was the best thing ever," ask whether you're valuing the experience or just its final frame.
- Watch for it being used on you. Suspiciously delightful endings — the surprise gift at checkout, the over-the-top farewell — may be engineering your remembered utility while your experienced utility quietly suffered. The countermove is the experiencing-self question, asked deliberately.
- Hold the humility. Apply the strong version (engineer the end, never finish on the low) confidently in short, designed experiences — that's where the evidence is solid. Apply the sweeping version (peak-end explains your whole year/marriage/career) with real skepticism — that's where the evidence thins out. Matching your confidence to the strength of the evidence is itself part of mastering this.
How It Connects
the Law of Relativity most of all — Relativity says nothing has fixed value, that everything is judged by comparison and frame, and the Peak-End Rule is Relativity operating inside a single experience across time: the same lived minutes get a wildly different value depending on which two moments your memory chooses as the frame, and how it ends resets the comparison point for everything before it. Also connects to Cause & Effect (your remembered verdicts become the causes of your future choices — so a distorted memory propagates into real decisions), to Rhythm (the trajectory/slope reading of the "end" effect is fundamentally about motion over time), and to Kahneman's broader dual-system work — the experiencing/remembering split is a close cousin of System 1 / System 2, two processors that don't share the same goals. Thinkers in the orbit: Daniel Kahneman and Amos Tversky (the heuristics-and-biases program it grew from), Donald Redelmeier (the clinical anchor studies), and Barbara Fredrickson (co-author of the cold-pressor work, later known for the broaden-and-build theory of positive emotion).
You don't remember your life — you remember its peaks and its endings, and then you live as if that highlight reel were the truth.