The peak–end rule

The peak–end rule predicts that your memory of an experience strongly correlates with the average of how you feel at the peak of the experience and the end of the experience. 

Have you ever thought about how strange it is that the moment you’re experiencing right now will never happen again?

The only thing you can take with you into the future are memories.

And your memories are influenced by feelings.

That’s right. Feelings. Not the content of the present moment, but your feelings in the present moment.

And what’s crazy is that most of those feelings don’t last, either.

So how, then, do you remember anything?

Today we’re going to look at the strange ways your brain uses feelings to reconstruct memory. Specifically, the feelings at two specific moments in an experience: the peak and the end.

The peak is the best moment of a pleasurable experience or the worst moment of a painful experience. The end is what you’re feeling right when the experience finishes.

Your brain relies on these two moments, the peak and the end, as a proxy for how you felt the rest of the time (now forgotten). This is what becomes the basis for your memory of the experience. Your brain does this in a predictable and reliable way. It’s called the peak–end rule.

Why the peak–end rule matters

There are three reasons the peak–end rule is so important to understand.

1. First, it matters because it reveals a systematic way you misremember. We’re going to look at that in detail, including a bunch of examples that show how it works, as well as a few examples that explain when it doesn’t work, and why that might be.
2. Second, it matters because you’re under the influence all the time of people who craft experiences for you. This could be as simple as the enjoyable experience you have when you eat out at a restaurant, or the experience you have of being manipulated when you view an advertisement, or the pain you experience when you’re about to have a medical procedure. We want to understand what others know about the peak–end rule so we know when we’re being influenced by it.
3. Third, the peak–end rule matters because it exposes the gulf between our experiencing self and our remembering self, and it raises the question: who’s in charge? (That’s right—you have more than one self, and they don’t always agree. This shouldn’t be too shocking: there’s probably one version of you who’s trying to refrain from eating too much dessert, and another version asking for more.)

We’re going to explore this in the rest of this post.

But first, we need to take a step back and define what we mean by “the present.” It sounds like it should be straightforward, but it isn’t. So let’s start there.

How you experience the present

The present, as you perceive it, is three seconds long. That’s how long “now” is. Something strange happens at the three-second mark for all people from all cultures. People recite poetry in a rhythm of three seconds, regardless of the language.¹ Infant hand gestures take place in groups that last around three seconds. Adults do the same thing: the hand motions people use when they talk are grouped in three-second clusters, too.²When you wave goodbye, shake someone’s hand,³ or hug someone,⁴ you do so for three seconds, on average. It takes about 3 seconds to divert your attention from one thing to another.⁵ The three-second mark acts as a boundary marker for human behavior.

You might object that you could categorize human behaviors in thousands of ways and that you could find just as many interesting boundary markers at two seconds or four seconds. You could accuse me of cherry-picking the behaviors which happen to coalesce around three seconds.

Fair enough. But there’s something more going on. Neuroscientists have found within a window of about 2.5 seconds, your brain perceives groups of stimuli as a single unit. But by around 5 seconds, it perceives them as separate. After three seconds, it gets harder to experience rhythms as single units, for example. 

Both neuroscientists and those who study human behavior have arrived at three seconds as “the present.” In a way, neuroscientists look at the way the brain perceives the present, while those who study human behavior look at the ways the brain’s perception of the present manifests itself in human behavior. And both arrive at the same place: the present, as we perceive it, functions as a kind of three-second moving average.

When you string a bunch of these presents together, you get an experience. Your commute to work. The movie you saw. The conversation you had with a friend. The meal at a restaurant.

But there’s a big problem. The experiences you have are made up of more moments than you can remember. If “the present” is three seconds long, then a minute-long experience contains twenty moments. An hour-long experience contains 1,200 moments. Go on a week-long vacation and you’ve got more than a hundred thousand. You might remember your twenty-minute commute, but you won’t remember the roughly 400 instances of “now” that you experienced as you drove. Most of these moments you will forget. With so many vanishing memories, how can we know the memories about our experiences are accurate? 

One answer might be that you simply recall the sum of the moments you had. This is the area shaded under the graph:

For your brain to calculate the sum of your experiences, it would still need to have all those moments stored away somewhere in your subconscious, readily available for a quick calculation. This doesn’t work.

Maybe instead of remembering an experience based on the sum of its moments, you remember based on the average of all the moments, like this:

We’ve already stated that you can’t remember every single moment of an experience. Since you can’t calculate your average feeling if you don’t know how you felt in each moment, we can safely say it’s not the average that produces a memory.

So if your brain doesn’t reconstruct an experience based on the sum or average of the combined moments, then how does it do it? If you can’t remember everything, how do you remember anything?

Here’s how.

First, your brain samples a few key moments from an experience. Just two, in fact. The first is the peak. It’s the moment of the greatest pleasure, the worst pain, the most intense drama, or the most boredom. The second is the end. It’s how you felt during the final moments of an experience. Next, your brain uses those sampled moments to guess at how you felt during the rest of the moments—the moments you’ve forgotten. Finally, using backward-looking inductive inference, your brain constructs a global retrospective evaluation of the experience.

This process of sampling the peak moment and end moment of an experience and using it as a proxy to guess at the rest of the experience is called the peak–end rule. The peak–end rule predicts that your memory of an experience strongly correlates with the average of how you feel at the peak of the experience and at the end of the experience. This rule holds regardless of the sum of how you feel, the average of how you feel, and even the length of the experience.

5 examples of the peak–end rule

Example #1: Getting a colonoscopy

Psychologists first became aware of the peak–end rule by studying colonoscopy patients. They recorded how much pain people felt during their procedure–which, at the time, was more painful than it is today. In the study, 154 patients indicated on a scale of 0 to 10 how much pain they were experiencing each minute. Then, one hour after the colonoscopy was finished, the researchers asked the same patients how much pain they remembered feeling. Finally, they followed up again one month later and asked patients to recall how painful their experience was.

The pain reports revealed three surprising findings:

1. First, psychologists could predict how much pain people remembered by averaging the peak pain and end pain while disregarding all other pain patients experienced.
2. Second, psychologists found no correlation between average pain, total pain, or even the duration of the experience. How long the procedure lasted didn’t matter. In other words, a 4-minute colonoscopy and a 67-minute colonoscopy with identical peak pain and end pain will be remembered as being equally painful, even though the 67-minute colonoscopy had a far greater total pain. 
3. Third, because the duration of the experience doesn’t matter, long colonoscopies could be remembered as less painful than short colonoscopies. When doctors extended the colonoscopy for a few minutes but made those extra minutes just a little less painful, the patients remembered the entire colonoscopies as being less painful, even though the total pain they experienced was greater.⁶

Example #2: Plunging your hand in cold water

The peak and end moments affect your memory in funny ways, but does this matter? Does it affect your future behavior? Psychologists tested this with a different experiment. Thirty-two people placed their hands in ice water at 14.1°C. This isn’t enough to freeze your hand or cause lasting damage, but it’s still uncomfortable. As with the colonoscopy experiment, psychologists measured discomfort in real–time. Everyone put their hands in the cold water twice. The first time, they placed their hands in the cold water for 60 seconds. The second time, they placed their hands in the water for 90 seconds–but during the final 30 seconds, the researchers raised the temperature of the water from 14.1°C to a slightly more comfortable 15.2°C—still very cold. As with the colonoscopy experiment, the researchers found 22 of the 32 people preferred the longer trial, even though the longer trial contained all the pain of the short trial and then some. This tells us that the memory of an episode—largely determined by the peak–end rule—predicts people’s future behavior more than the actual, real–time experience of that episode.

Not only do your remembered feelings not match your experienced feelings, but it’s your remembered feelings that predict your future behavior. When researchers asked the participants in the ice water experiment about their memory of it, the participants got it wrong in three ways:

1. First, they claimed the long trial was more comfortable. This memory contradicted their real–time experience: the long trial contained all the discomfort of the short trial, and then some. 
2. Second, they said the long trial wasn’t as cold at the most extreme moment. But that wasn’t true either: both were exactly 14.1 °C at their most extreme. 
3. Third, and most incredibly, the better ending of the long trial convinced 18.75% of the participants that the long trial was shorter. A short bad experience will be remembered as longer than a long bad experience with a slight improvement at the end.⁷

This last finding tells us that an incorrect memory of an experience can warp time.

Example #3: Studying

Educators have unwittingly used the peak–end rule to help their students learn more by altering their study habits. Students are more likely to study if they had a good experience the last time they studied. Students with favorable memories of past study sessions are more likely to study again, compared to students with negative or neutral memories of past study sessions. Just like colonoscopies and ice buckets, the peak–end rule predicts that students will have a more enjoyable memory of a long study session that ends well compared to a short study session.

Bridgid Finn, a psychologist who studies learning and memory, found that English-speaking students liked learning long lists of Spanish words more than short lists. But this was true so long as the long lists ended with a few less challenging words. Her experiment showed that, while people got a better grade after studying the short list, they preferred the long list. In fact, 70% of students thought the long list was easier–even though it was not only harder but also caused them to perform worse on the recall test.⁸ A better ending, averaged with an identical peak, produced a more favorable memory, even though both the overall real–time experience and the grades were worse.

But wouldn’t a worse score on the test discourage students from studying more? According to other research Finn and her colleagues have conducted, the desire to study comes not from test performance but from the remembered enjoyment of previous study sessions. This is why a longer study session with a better ending will encourage future learning, even if it doesn’t produce higher grades right away.⁹ The optimal scenario, though, involves a study session that ends on a high note and a test that ends on a high note, too.¹⁰

Example #4: User interface design

The peak–end rule predicts how they interact with design elements on your website, too. As with colonoscopies, ice bucket plunges, and studying, when people interact with the interfaces you design, they are doing so in time. (Designers spend a lot of time thinking about the second and third dimensions; the fourth dimension is just as important.) Like all experiences in time, these interactions contain peaks and ends—the moments that predict how people remember their experience with your brand. (Or, in some cases, like checkout forms, whether you keep or lose the customer.)

Researchers discovered the peak–end rule applies to user interface design by creating a mock online form. Think about forms you’ve encountered: you type information in a few fields and then hit next. Then you type some more, and hit next again until you’ve completed the form. In the experiment, people were asked to set 25 sliders to a number between 1 and 50. The sliders were spread out over 5 pages. Once the sliders on a page had been set, subjects clicked “next” at the bottom to proceed to the next page. Although everyone worked through 25 sliders, not everyone did so in the same way:

• Some pages had lots of sliders, while others had only a few.
• Sometimes, the pages with lots of sliders appeared near the beginning of the sequence, while other times, the pages with lots of sliders appeared near the end.
• Some sliders were large and easy to move, while others were smaller and more difficult to move.

After people worked through their pages of sliders, researchers asked them how much they enjoyed the experience. Not surprisingly, people preferred a series that had a positive peak early on, followed by an improving trend at the end. For example, they preferred a series like 7-2-7-6-3 to 3-8-4-4-6.

As you can see, the peak–end rule applies to interface design. When users experience your design elements in time, they’ll remember them favorably if they have a positive peak and a positive end.¹¹

Example #5: Advertising

People respond better to ads with positive peaks and ends, too. Researchers asked 27 people to view 30 ads—a total of 810 views—and provide real–time ratings as they watched them. The ads ranged in length from 30 seconds to 90 seconds. The researchers found close correlations between how much people liked the ad at the peak and the end and how much they liked the ad after watching it.

But researchers discovered something else interesting about how peaks and ends affected viewers’ later memory. They found viewers were more likely to remember the ad positively when the peak appeared early rather than later commercial. They also found that as more time elapsed after the peak, people liked the ad less.¹²

This advertising study hints at something else about peaks and ends: you can have two experiences with identical peaks and ends, but you will remember the experiences differently depending on several specific features of those peaks and ends. We’ll explore this next.

8 factors that determine how peaks and ends influence your memory

In this section, we’re going to explore eight conditions where peaks and ends can have a greater effect on how an experience is remembered:

1. The recency of the peak: whether the peak occurs near the start or end of the experience affects memory.
2. The number of peaks: if there are two or more equal peaks, does the peak–end effect get stronger or weaker?
3. The duration of the peak: do peaks that last a long time have more effect than peaks that are short or instantaneous?
4. The trend of the experience: does an improving or worsening trend affect how peaks and ends factor into memory?
5. The rate of change: do quick ramp-ups to the peak moment affect memory?
6. The ending trend: does an improving or worsening trend at the end affect how the end is remembered?
7. Relative duration: in what kinds of experiences does duration interact with the peak and the end?
8. Adding a better beginning: all things equal, how does adding a better ending compare to adding a better ending?

Let’s dive in.

1. The recency of the peak

Does the recency of the peak affect how much the peak influences your memory of an experience? Yes—but only a little. In a study conducted by Charles Schreiber and Daniel Kahneman at the University of California, Berkeley, participants listened to a series of annoying sounds at 60 to 80 decibels. They indicated in real–time how unpleasant the sounds were on a scale of -10 (very unpleasant) to 0 (neutral) to 10 (pleasant). Then they rated how unpleasant the overall experience was. When Schreiber and Kahneman compared the real–time rankings to the overall rankings, they found that the recency of the peak mattered, but only a little. For a negative experience, like hearing a loud sound for an extended period, people have a slightly better memory when the peak occurs early.¹³

Summary: all other things being equal, a more recent peak has more impact on memory than an early peak, but only a little.

2. The number of peaks

What if there are two equal peaks? Does the peak–end effect get stronger or weaker? In the same study of annoying sounds, Schreiber and Kahneman had some students hear the sound at its loudest only once. But for other students, they had them hear it twice, producing two equally negative experiences. Once again, Schreiber and Kahneman compared the real–time rankings and the overall rankings and found that a second peak didn’t change the overall memory of the experience.¹⁴

Summary: all other things being equal, adding a second, equal peak doesn’t change your memory.

3. The duration of the peak

Do peaks that last a long time have more effect than peaks that are short or instantaneous? Schreiber and Kahneman found that when the most annoying sound lasted longer, it produced a slightly more negative memory of the entire experience. 

At first glance, this seems to contract the earlier claim that duration doesn’t matter. Recall that people remember a short colonoscopy and a long colonoscopy as being equally uncomfortable so long as their peaks and ends are the same. However, while the duration of the overall experience doesn’t matter, the duration of the peak does matter.¹⁵

Summary: all other things being equal, a long peak has slightly more influence over memory than a short peak.

4. The trend of the experience

Whether the overall trend of an experience is getting better or worse affects how peaks and ends factor into overall memory. In a 1992 experiment, psychologists Carol Varey and Daniel Kahneman compared how people responded to experiences with the same peaks and ends. In one version, the overall trend was getting worse, while in the other version, the overall trend was getting better. People gave the worsening trend a 13% lower rating than its opposite. 

Varey and Kahneman found an even bigger difference when the trendline was steeper. People rated a 4-to-6 series (bad to good) 5.3% better than a 6-to-4 (good to bad) series, but they rated a 2-to-8 (very bad to very good) series 19.1% better than an 8-to-2 series (very good to very bad).¹⁶ 

Another study revealed something even more interesting. When the experience is pleasurable and the trend is improving, people want the experience to be short. But when the experience is painful and the trend is worsening, people want the experience to be long.¹⁷

Summary: all other things being equal, an improving trend produces a better memory than a worsening trend.

5. The rate of change

Whether you get to the peak quickly or slowly affects how much the peak affects your overall memory. In a 1998 study, Dan Ariely found people remember an experience as more painful when the peak pain is reached quickly.¹⁸ And another study of advertising found that people liked television commercials more when they had a steep, quick ramp-up to the peak moment, rather than a slow increase.¹⁹ Third, in the study of annoying sounds I mentioned earlier, people gave a worse overall ranking when the volume went up quickly, compared to when the volume went up slowly to the same level.²⁰

Summary: all other things being equal, a quick ramp-up to the peak causes the peak to be a greater factor in the memory of the experience.

6. The ending trend

We’ve already seen that an improving trend produces a better memory than a worsening trend. It turns out that the trend right at the end affects how much the ending is factored into the peak–end rule. When the trend is improving at or near the end, people remember an experience more positively overall. In a 1991 study, Carol Varey and Daniel Kahneman asked people to rank several painful sequences. They found that a pain rating of 5 is remembered as less painful if the ending trend is improving, compared to the same rating when the ending trend is worsening. For example, A 2-5-8 sequence was rated a 64, while a 2-5-8-4 sequence was rated a 53. An 8-7-6-5-4-3-2 sequence was rated 46.4, while an 8-7-6-5-4 sequence was rated a 53.9. When the pain was worsening at the end of the experience, people remembered it as more painful, but when pain was decreasing at the end of the experience, people remembered it as less painful—even though they experienced more total pain overall.²¹

Compare the trajectory of two hypothetical days:

1. You drive your car to work, park your car, get out, and hear a loud scraping noise. When you turn around, you see that one of your co-workers hit your car as they pulled into the spot next to yours. You’ll need to spend $175 to get it fixed. Not a great start to your day.
2. Now, imagine a similar scenario. Your co-worker hits your car, just as before, but this time, it’s worse: it will cost you $200 to fix it. Fortunately, a few minutes later when you get inside, you notice your boss has placed a $25 Starbucks card on your desk as a thank you for landing a new client.

Which scenario is worse? Most people say the first one.²² But why? Both scenarios are economically equivalent: you’re out $175. The reason the second scenario feels better is because it represents an improving trend. In a 1991 study, researchers gave people a set of similar hypothetical scenarios:

• Scenario A: first lose $85, but then win $80, for a net loss of $5
• Scenario B: lose $5
• Scenario C: first lose $15, but then win $85, for a net gain of $70
• Scenario D: first win $85, but then lose $15, for a net gain of $70

As you can see, A and B are economically equivalent: in both cases, you lose $5. And C and D are economically equivalent, too: you win $70. Yet 73% of people who participated in the study preferred the scenarios that were on a positive trajectory, while only 14% preferred scenarios that began positively and ended badly—even though the scenarios were economically equivalent.²³

Trends matter in user experience design, too. Think about the last time you tried to open a webpage on a phone with a poor signal. The bar fills in from left to right, but not at a constant pace.

People like progress bars that accelerate more than progress bars that move a constant rate. In a study, people compared 445 sets of progress bars—990 progress bars total. It took 5.5 seconds for all progress bars to fill in. People thought progress bars loaded more slowly when they started fast and slowed down, and they thought progress bars loaded more quickly when they started slow and sped up. In fact, in another example of a time warp, people perceived progress bars that sped up toward the end as taking less than the actual 5.5 seconds to load.²⁴

A well-designed progress bar with an improving ending trend warps your experience of time and makes you think it loaded faster than it really did.

Summary: all other things being equal, an improving trend at the end improves the memory of the experience overall.

7. Relative duration

We’ve already seen that duration doesn’t matter—long colonoscopies and short colonoscopies produce memories of equal discomfort so long as they have the same peaks and ends. However, duration can sometimes make a difference in your overall memory when experiences have greater amounts of intensity.

First, let’s establish how duration works in zero-intensity-change experience. In a study, people placed their arm on an element heated to 45°C (roughly 95°F)—painful enough to be very uncomfortable, but not so painful as to cause lasting harm. One group touched the heating element for 10 seconds, and another group touched it for 14 seconds. Both groups remembered feeling the same amount of pain, even though the group who touched it for 14 seconds experienced more pain overall.²⁵ Even for an experience without peaks and ends, duration doesn’t matter.

Second, let’s look at how duration works when the intensity changes. In a different experiment, people ranked their discomfort at 5-minute intervals. A 15-minute series of discomfort levels of 2-5-8 was rated a 64.0, while a 35-minute series of discomfort levels of 2-3-4-5-6-7-8 was rated a 68.2—a little worse. 

In another version of the same experiment, researchers found that the 30-minute series of discomfort levels of 2-2-4-4-6-6 was evaluated as 2.8% worse than the 15-minute series of discomfort levels of 2-4-6. Through these and similar experiments, researchers found that doubling the length of experienced pain increased the remembered pain by 3.5%.²⁶

Summary: all other things being equal, while the peak–end rule is still primarily responsible for the way an experience is remembered, the duration can matter in specific instances, even if only a little.

8. Better beginnings

We haven’t touched on beginnings yet. Specifically: all things being equal, how does adding a better beginning compare to adding a better beginning?

Sometimes, the ending doesn’t matter at all. (Wait! Doesn’t that contract everything you’ve been saying so far? Well, sort of. But stay with me…) In a 2016 study, Stephanie Tully, from Stanford, and Tom Meyvis, from NYU, took a great experience at the end and slotted it in the middle. In one experiment, they watched 750 people run a fun-run-style obstacle course. After the race, Tully and Meyvis asked people how much fun they had on a scale of 1 to 10. They also asked people how much they liked or didn’t like each obstacle on a scale of 1 to 5. As predicted by the peak–end rule, the obstacle at the end was a predictor of how people felt about the race as a whole. But here was the surprise: so did many of the other obstacles. In this race, the ending mattered, but not more than any other moment.²⁷

Other times, a great experience at the end can be moved to the beginning and produce the same overall memory. In another experiment, Tully and Mayvis had people read a series of cartoons and rate how funny they were. They also asked people whether they would, hypothetically, be willing to pay $1 for more cartoons similar to the ones they had just read. Tully and Mayvis found that people rated bad cartoons as worse than good cartoons, which wasn’t a surprise. They also found bad cartoons decreased the overall enjoyment of the series when they were placed last. No surprise there, either: that’s the peak–end effect at work. 

Or was it? 

Tully and Mayvis also found bad cartoons decreased the overall enjoyment of the experience in an equal amount when they were placed first in the series.²⁸ 

Summary: better endings produce a positive memory, but so do better beginnings (sometimes).

In these two experiments, a positive ending doesn’t seem to matter. What is it about these experiences, unlike all the others we’ve looked at so far, that makes the peak–end rule a poor predictor of memory?

Every rule has exceptions, and the peak–end rule has four. Let’s look at each of these.

Exceptions to the peak–end rule: 4 cases when it doesn’t work

There are four kinds of experiences where the peak–end rule can’t predict your memory very well:

1. When you recall your experiences using semantic memory instead of episodic memory (we’ll explain what semantic memory and episodic memory are in a moment)
2. When you’re performing goal oriented tasks or having experiences with defined, known endings
3. When you’re doing something with a pre-defined sequence
4. When you’ve made a strong prediction or you have an expectation about how an experience will go

Let’s take a look at each of these.

Exception #1 to the peak–end rule: The passage of time

The peak–end rule fades with time. In 2013, researchers wanted to understand how people remembered vacations. In their study, they tracked 72 people on vacation. At the end of each day, people reflected how happy they were on a scale of 1 (very unhappy) to 7 (very happy). After the vacation was over, they were asked to give the same rating. One group gave their overall ratings a day after, 3 weeks later, and 7 weeks later. Another group gave their ratings a day after, 7 weeks after, and 11 weeks after. At one week after and three weeks after, the researchers could predict the remembered experience using the peak–end rule. But at 7 weeks, something started happening: the peak–end rule became a less reliable predictor of memory—and it became less so as more time passed.²⁹

Episodic memory vs. Semantic memory

Like all memories, peak memories and memories fade over time, and they do so at a fairly predictable rate.³⁰ That’s because, at first, your brain relies on episodic memory, or the memories from the specific episode. Then, as time passes, your brain shifts to relying on semantic memory, or inferences from generalizations about the kind of experience. For example, let’s say you go on a scary roller coaster. In the weeks following the ride, your memory will be informed by the peak and end moments from that specific experience. But months or years later, your memory will be mostly informed by what you know about scary roller coaster rides in general. This is a neat shortcut because it’s almost certain that your experience of the roller coaster ride matches what roller coaster rides are like in general. Your memory isn’t wrong, per se. It’s just based on a source less tied to a specific experience.

Psychologists have proposed one reason our brains accept this trade-off is that sometimes we need to learn something quickly and remember how we felt about it. Episodic memory does that well. But we also need to have memories that are stable and transcend episodes—to arm, prepare, and equip us for new experiences. Semantic memory does that well. Plus, to learn new things, we need to forget others—our brain’s’ memory capacity is not infinite.³¹ (This is similar to how memory caching works on your computer.)

The memory of pregnancy and birth

Women who have given birth to a second or third child show evidence of this transition from episodic to semantic memory. Because they’ve already given birth once before, they have a more defined semantic category of “giving birth,” which means they’re quicker to transition from episodic memory to semantic memory after having a second or third child. Women who have given birth to only one child have a less defined semantic category, which means their episodic memory lasts longer. A 2014 study verified this. Researchers monitored 320 women from the beginning of labor to the birth of their child. Every 20 minutes, they were asked to indicate their pain on a scale of 0 (no pain) to 100 (worst pain imaginable). Then, after they had given birth, the researchers asked the women to indicate how much pain they remembered feeling. Two days after delivery, the peak–end rule predicted the amount of recalled pain. (As expected from the peak–end rule, there was also no correlation between delivery duration and the amount of recalled pain. It seems preposterous that a 24-hour labor is remembered as no worse than a 3-hour labor, but that’s what women are reporting.) 

However, something strange happened at the two-month mark. At two months, the peak–end rule predicted remembered pain, but only for mothers who had never given birth before. For mothers who had already had a child, the peak–end rule could not be used to predict their recalled pain. 

This confirms that the shift from episodic memory to semantic memory took longer for women who hadn’t given birth before. But second- and third-time mothers had already developed a semantic category for “childbirth”; their recall of their most recent childbirth could more quickly transition from episodic memory—the memory of the specific experience—to semantic memory—the memory of the category of “childbirth” as a whole.³²

Exception #2 to the peak–end rule: Goal-oriented experiences

The peak–end rule also does not predict memory when the experience is goal-oriented. Consider two examples of goal-oriented experiences:

1. Standing in line. Whether you’re standing in line at the grocery store or at the airport, the goal of standing in line is to get out of line.
2. Looking for something you’ve lost. You’ve probably heard people say, after finding something, “That’s the last place I looked!” Of course: why would you keep looking for the item after you’ve found it? (This is why you’ve never heard someone say: “I found it in the second-to-last place I looked!”)

The peak–end rule doesn’t work for goal-oriented experiences because the peak is the end. Peaks and ends collapse into a singular moment that occurs right when the experience ends.

The peak–end rule also doesn’t work because you know it’s coming. In a study, researchers simulated the act of waiting in line. They measured people’s feelings while they waited on a scale of -10 to 10. Some people stood in short lines that moved slowly, while others were long and moved quickly. The pace people moved through the lines varied, too: some lines accelerated, while others slowed. The catch was that everyone stood in line for the same amount of time, regardless of how quickly the lines moved (or didn’t move). People who stood in short lines indicated a peak of 5.0, while people who stood in long lines indicated a peak of -8.4. No surprise there: people hate standing in lines; if you’re in a line that feels short, you’ll feel happier while you’re in line.

What was surprising was that feelings of peak annoyance while standing in line did not affect how people remembered their experience later. Only the feelings at reaching the end of the line mattered. And people’s feelings at the end of the line were largely affected by only one factor: how quickly the line moved.³³ For goal ending experiences like standing in line, peaks do not factor in your memory of the experience.

Exception #3 to the peak–end rule: Experiences with pre-defined sequences

The peak–end rule also doesn’t predict your memory for experiences that have an already-defined sequence. They lack an element of surprise. You know what to expect. And this means you’re less likely to rely on a peak moment or an end moment as a cue for remembering the experience as a whole.

Sometimes sequences are defined by cultural norms. Take the way we end conversations, for example. If you’ve ever seen The Simpsons, you’re probably familiar with the character Apu, who owns a convenience store. In every scene Apu appears in, his final words are “Thank you, come again,” illustrating the ritualistic way a store clerk would close a conversation. However, there’s one scene where Homer Simpson, the main character, has an altercation with Apu, and Apu kicks him out. As Homer leaves, Apu tells him, “Mr. Simpson, get the hell outta my store. Please come again.” The phrase “please come again,” while technically positive, functions purely as a social convention, just like “see you later.” 

Even though phrases like “please come again” and “you, too” function as positive endings to conversations, people do not rely on them to form memories of the experience because their meaning isn’t in their substance. In a 2004 study, researchers evaluated 97 service calls for a large financial services provider. Calls lasted 208 seconds on average. The researchers noticed that the peak–end rule didn’t work when people were later asked how their calls went. That’s because there’s an established social protocol for how these kinds of calls should end: with a “have a nice day” or “thank you for your time.” The peak of the call could still be used to predict how the call was remembered, but not the end.³⁴

Exception #4 to the peak–end rule: Experiences that have been predicted or that have strongly-held expectations

Sometimes, your memory of an experience is affected more by your expectations going into the experience than the peaks and ends from the experience itself. This is because you see what you expect to see. Take, for example, a now-famous video of a man in a gorilla costume wandering through a game of basketball. Between one-third and one-half of people who watch this video don’t see the gorilla. That’s because they’ve been asked to count how many times the players pass the ball. Because they’re focused on looking for one thing, they miss the gorilla altogether.³⁵

You approach experiences in the same way. If you expect to have a good time, you probably will. If you expect not to, you won’t. In a 1994 study, Kristen J. Klaaren and her colleagues showed a film to two groups of people. The first group was told the film would be great. The second group wasn’t told anything. Which group liked it more? You can probably guess: the group who was told they would enjoy the film.³⁶

In another study, 38 students went biking for three weeks. Researchers asked them to indicate how they felt about the experience at 12 intervals: twice before the trip, eight times during the trip, and twice after the trip. The results indicated that peoples’ memories of the trip were dictated more by their expectations going into the trip than their actual experience while on the trip. The authors of the study concluded that people “enjoyed the trip less than they anticipated and less than they later remembered.” But how much? Before the trip, only 5% of the bikers anticipated that they would feel disappointed or frustrated during the trip. During the trip, 61% of the bikers reported feeling disappointed or frustrated. But after the trip, only 11% remembered feeling disappointed or frustrated. Moreover, there was a close association between people who expected to be disappointed and remembered being disappointed. And people who expected to have fun remembered having fun—even though most of them experienced disappointment, too.³⁷

The same thing happens when you go on vacation. Researchers tracked 46 people before, during, and after they took a trip. People were asked to indicate how positive or negative they felt, and asked to indicate their agreement with statements like “I expect to enjoy spring break” and “I will be satisfied with this vacation.” As with the bike trip, people enjoyed their vacations less than they predicted and less than they remembered. The biggest predictor of whether they would repeat the vacation was their remembered experience; the second-biggest predictor was the predicted experience; and the third-biggest predictor was the actual experience.³⁸

What both the bike trip study and the vacation study tell us is that, while memory is largely determined by peaks and ends, sometimes the peaks and ends are completely overruled by the predictions going into the experience.

Who’s in charge: The experiencing self or the remembering self?

As we have seen, the peak–end rule reveals that your experienced feelings don’t match your remembered feelings. Your experiencing self and your remembering self see the world in different ways. And these differences are not small: one likes long colonoscopies; the other likes them short. One prefers a short childbirth; the other doesn’t care. One likes watching short advertisements; the other likes them longer. And so on.

Though you have an experiencing self and a remembering self, your two selves are still you: the long-colonoscopy-preferring version of you and the short-colonoscopy-preferring version of you are the same person. 

The question is: who’s in charge? Which one is really you? Who’s calling the shots, the remembering self or the experiencing self?

Let’s answer this with an illustration. Pretend you’re about to go on vacation. But there’s a catch: you can’t take pictures. And that’s not all. You won’t remember a single thing from your vacation. The moment you return home someone will hold a device to your head that will wipe your memory. You won’t even know you went on vacation.

Would you still go? 

At the core of this hypothetical situation is the question: do you go on vacation to experience it, or to remember it? The experiencing self would go again; it’s your remembering self who’s not so sure.³⁹

A few years ago, I did this very thing (without the memory-wiping guy at the end). I spent three weeks traveling in the Middle East without a camera. Why? Because I didn’t want my desire to take pictures to impede my ability to enjoy what I was seeing and doing. I didn’t want my remembering self to ruin the moment my experiencing self was having. When I tell people this, they’re incredulous that I don’t have any documentation of my trip—not a single photo. And it’s true, I would probably have a better memory of the trip had I taken pictures. But I don’t regret it. (Plus, all the places I visited have been documented by photographers far more skilled than me.)

In short, you go on vacation to make memories. The reason for your trip is not the experience you’ll have, but the memories you’ll make. In cases like this, your experiencing self is thus subordinate to your remembering self.

But what about negative experiences? Pretend that, instead of going on vacation, you’re about to undergo a painful medical procedure. Would you rather your experiencing self feel more pain so your remembering self feels less?

Or what about prison sentences? Should we demand lengthy prison sentences for heinous crimes when a shorter prison sentence with the same peaks and end will produce the same result as the long one?

Should teachers encourage students to quit studying while they’re ahead, so students’ remembering selves view the study session more positively and are more likely to study again—even if their grades take a short term hit?

To answer this, let’s return to the claim I made at the beginning. I showed you that what you experience as “the present” lasts roughly three seconds. You will never experience that moment again, and your memory of it will degrade to zero almost immediately. 

That’s the difference between the experiencing self and the remembering self. The experiencing self, while being the life you’re living right now, immediately disappears. It’s the remembering self, in the words of Daniel Kahneman and Jason Riis, that is “relatively stable and permanent. . . . The only perspective that we can adopt as we think about our lives is therefore that of the remembering self.”⁴⁰

Who you are isn’t just who you are right now. Who you are is also a crudely-calculated moving average of feelings you’ve had during your most important moments combined with your best attempts at salvaging what you can from the world as you experience it.

1. Turner, Frederick & Pöppel, Ernst. (1983). “The neural lyre: Poetic meter, the brain and time.” Poetry, 142. 277-309.
2. Trevarthen, C. (1999). Musicality and the intrinsic motive pulse: evidence from human psychobiology and infant communication. Musicae Scientiae, 3(1_suppl), 155–215.
3. Schleidt M. (1988). “Universal time constant operating in human short-term behavior repetitions.” Ethology 77, 67–75.
4. Nagy, E. (2011). “Sharing the moment: the duration of embraces in humans.” Journal of Ethology 29(2), 389–393.
5. Pöppel, E. (1978). “Time perception.” In: Held R, Leibowitz HW, TeuberHL (eds) Handbook of sensory physiology, vol 8. Springer,Heidelberg, 713–729.
6. Redelmeier, D. A., & Kahneman, D. (1996). “Patients’ memories of painful medical treatments: Real-time and retrospective evaluations of two minimally invasive procedures.” Pain, 66, 3–8.
7. Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A. (1993). “When more pain is preferred to less: Adding a better end.” Psychological Science, 4, 401–405.
8. Finn B. (2010). “Ending on a high note: adding a better end to effortful study.” Journal of experimental psychology. Learning, memory, and cognition, 36(6), 1548–1553.
9. Finn, B., & Metcalfe, J. (2008). “Judgments of learning are influenced by memory for past test.” Journal of Memory and Language, 58(1), 19–34.
10. Finn, B., & Miele, D. B. (2016). “Hitting a high note on math tests: Remembered success influences test preferences.” Journal of Experimental Psychology: Learning, Memory, and Cognition, 42(1), 17-38.
11. Cockburn, A., Quinn, P., & Gutwin, C. (2015). “Examining the peak-end effects of subjective experience.” Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, 357-366.
12. Baumgartner, H., Sujan, M., & Padgett, D. (1997). “Patterns of affective reactions to advertisements: The integration of moment-to-moment responses into overall judgments.” Journal of Marketing Research, 34(2), 219–232.
13. Schreiber, C. A., & Kahneman, D. (2000). “Determinants of the remembered utility of aversive sounds.” Journal of Experimental Psychology: General, 129(1), 27-42.
14. Schreiber, C. A., & Kahneman, D. (2000). “Determinants of the remembered utility of aversive sounds.” Journal of Experimental Psychology: General, 129(1), 27-42.
15. Schreiber, C. A., & Kahneman, D. (2000). “Determinants of the remembered utility of aversive sounds.” Journal of Experimental Psychology: General, 129(1), 27-42.
16. Varey, C. A., & Kahneman, D. (1992). “Experiences extended across time: Evaluation of moments and episodes.” Journal of Behavioral Decision Making, 5, 169–185.
17. Hsee, C. K., & Abelson, R. P. (1991). “Velocity relation: Satisfaction as a function of the first derivative of outcome over time.” Journal of Personality and Social Psychology, 60(3), 341-347.
18. Ariely, D. (1998). “Combining experiences over time: The effects of duration, intensity changes and on-line measurements on retrospective pain evaluations.” Journal of Behavioral Decision Making, 11, 19–45.
19. Baumgartner, H., Sujan, M., & Padgett, D. (1997). “Patterns of affective reactions to advertisements: The integration of moment-to-moment responses into overall judgments.” Journal of Marketing Research, 34(2), 219–232.
20. Schreiber, C. A., & Kahneman, D. (2000). “Determinants of the remembered utility of aversive sounds.” Journal of Experimental Psychology: General, 129(1), 27-42.
21. Varey, C. A., & Kahneman, D. (1992). “Experiences extended across time: Evaluation of moments and episodes.” Journal of Behavioral Decision Making, 5, 169–185.
22. Thaler, R. (1985). “Mental accounting and consumer choice.” Marketing Science 4(3), 199–214.
23. Ross, W. T. & Simonson, I. (1991). “Evaluations of pairs of experiences: A preference for happy endings.” Journal of Behavioral Decision Making 4, 273–282.
24. Harrison, Chris & Amento, Brian & Kuznetsov, Stacey & Bell, Robert. (2007). “Rethinking the progress bar.” UIST: Proceedings of the Annual ACM Symposium on User Interface Software and Technology.
25. Ariely, D. (1998). “Combining experiences over time: The effects of duration, intensity changes and on-line measurements on retrospective pain evaluations.” Journal of Behavioral Decision Making, 11, 19–45.
26. Varey, C. A., & Kahneman, D. (1992). “Experiences extended across time: Evaluation of moments and episodes.” Journal of Behavioral Decision Making, 5, 169–185.
27. Tully, S., & Meyvis, T. (2016). “Questioning the end effect: Endings are not inherently over-weighted in retrospective evaluations of experiences.” Journal of Experimental Psychology: General, 145(5), 630-642.
28. Tully, S., & Meyvis, T. (2016). “Questioning the end effect: Endings are not inherently over-weighted in retrospective evaluations of experiences.” Journal of Experimental Psychology: General, 145(5), 630-642.
29. Geng, X., Chen, Z., Lam, W., & Zheng, Q. (2013). “Hedonic Evaluation over Short and Long Retention Intervals: The Mechanism of the Peak-End Rule.” Journal of Behavioral Decision Making, 26(3), 225-236.
30. Rubin, D. C., & Wetzel, A. E. (1996). “One hundred years of forgetting: A quantitative description of retention.” Psychological Review, 103, 734– 760.
31. Robinson, M. D., & Clore, G. L. (2002). “Belief and feeling: Evidence for an accessibility model of emotional self-report.” Psychological Bulletin, 128(6), 934-960.
32. Chajut, E., Caspi, A., Chen, R., Hod, M., & Ariely, D. (2014). “In pain thou shalt bring forth children: The peak-and-end rule in recall of labor pain.” Psychological Science, 25(12), 2266–2271.
33. Carmon, Z. and Kahneman, D. (1995). “The experienced utility of queuing: Real-time affect and retrospective evaluations of simulated queues.” Ph.D. Thesis. Duke University: Durham, NC, USA.
34. Verhoef, P., Antonides, G., and de Hoog, A. (2004). “Service encounters as a sequence of events.” Journal of Service Research, vol. 7, no. 1, 2004, pp. 53–64.
35. Simons, D. J. and C. F. Chabris (1999). “Gorillas in our midst: sustained inattentional blindness for dynamic events.” Perception 28(9), 1059–74.
36. Klaaren, K. J., Hodges, S. D., & Wilson, T. D. (1994). “The role of affective expectations in subjective experience and decision making.” Social Cognition, 12, 77–101.
37. Mitchell, T.R., Thompson, L., Peterson, E., & Cronk, R. (1997). “Temporal adjustments in the evaluation of events: The ‘rosy view.’” Journal of Experimental Social Psychology, 33, 421–448.
38. Wirtz, D., Kruger, J., Scollon, C. N., & Diener, E. (2003). “What to do on spring break?: The role of predicted, on-Line, and remembered experience in future choice.” Psychological Science, 14(5), 520–524.
39. I’m borrowing this hypothetical situation from Daniel Kahneman, who has raised it in multiple interviews.
40. Kahneman, D., & Riis, J. (2005). “Living, and thinking about it: Two perspectives on life.” In F. Huppert, N. Baylis, & B. Keverne (Eds.), The science of well-being, 285–304, Oxford, England: Oxford University Press.