> ## Documentation Index
> Fetch the complete documentation index at: https://meta.niceshare.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Duration Neglect

> Duration Neglect is little or no independent effect of an episode's length on its later global rating. Film clips and a cold-water choice show the drop.

<Info>
  **Category**: Effects<br />
  **Type**: Little or no independent effect of an episode's length on a later global evaluation, once the affect of selected moments is held fixed<br />
  **Origin**: Barbara L. Fredrickson and Daniel Kahneman, *Journal of Personality and Social Psychology* 65(1), pages 45–55. The copy read here does not print a month. The record is July 1993. DOI 10.1037/0022-3514.65.1.45. Received 3 August 1992, revision received 7 January 1993, accepted 27 January 1993. Fredrickson at Duke University; Kahneman at the University of California, Berkeley<br />
  **Also known as**: the neglect of duration, their phrase for giving length little weight in that later rating
</Info>

<Note>
  **Quick Answer** — Duration Neglect is little or no independent effect of an episode's length on a later global rating of that episode, once the affect at selected moments is taken into account. Fredrickson and Kahneman named it in 1993 with plotless film clips. In a separate cold-water session the same year, 22 of 32 men chose to repeat the 90-second trial rather than the 60-second trial. The longer trial contained every cold second of the shorter one, and then 30 seconds more.
</Note>

## What is Duration Neglect?

Duration Neglect is little or no independent effect of an episode's length on a later global evaluation of that episode.

> Weighted averaging, in contrast, entails duration neglect, which we define as little or no independent effect of duration on retrospective evaluations of affective episodes.

That sentence is Fredrickson and Kahneman's definition. A global evaluation here is one score for the whole episode, given after it ends. The length can be noticed and still carry almost no separate weight. An everyday picture is a meal scored by the spiciest bite and the last bite. The number of bites can stay off the score.

[Peak-End Rule](/effects/peak-end-rule) is the companion claim that the worst or best moment and the final moment carry the weight. Duration Neglect is the missing weight on length. It is not a second name for those two moments. [Focusing Illusion](/effects/focusing-illusion) is a judgment pulled toward one salient part of a life or a scene. Duration Neglect is narrower. The part left out is the clock on one affective episode.

### Duration Neglect in 3 Depths

* **Beginner**: Thirty-two men each held a hand in cold water twice. One trial lasted 60 seconds. The other lasted 90 seconds and ended a little less cold. Twenty-two chose the longer trial for a repeat. The extra half-minute was still uncomfortable.
* **Practitioner**: Before you treat two memories as equal, write the length beside the rating. In the film study, people judged the long clips to be about twice the short ones. Once peak affect and end affect were held fixed, the partial link from length to the global rating was about zero.
* **Advanced**: A raw link between length and the rating is not the test. On the aversive films, discomfort grew as the clip continued, so the immediate global rating did move with length. That movement was accounted for by the change in the moment-to-moment affect. The partial correlation of length, after peak and end, did not remain.

## Origin

The name is the title of the 1993 film paper. The cold-water choice, the profile of another person's discomfort, and the later colonoscopy counts are separate reports. They do not share one sample.

Barbara L. Fredrickson and Daniel Kahneman published "Duration neglect in retrospective evaluations of affective episodes" in the *Journal of Personality and Social Psychology*, volume 65, number 1, pages 45–55. The copy read here does not print a month or a DOI. The record is July 1993, DOI 10.1037/0022-3514.65.1.45. The manuscript was received on 3 August 1992, the revision on 7 January 1993, and it was accepted on 27 January 1993. Fredrickson was in the Department of Psychology: Social and Health Sciences, and in the Center for the Study of Aging and Human Development, at Duke University. Kahneman was in the Department of Psychology at the University of California, Berkeley. Correspondence addresses in the paper are Duke, Box 90085, Durham, North Carolina 27708-0085, and 3210 Tolman Hall, Berkeley, California 94720.

They define the term against temporal integration. Integration would add the affect of each moment, so a longer bad episode would be worse, and added pleasant moments would make an episode better. Weighted averaging of a few snapshots does not do that. Length then has little or no independent effect. They call that Duration Neglect. The snapshots they test are the peak, the most extreme moment, and the end, the mean of the final 10 seconds on the film meter.

An earlier profile study is not this film test. Kahneman, Fredrickson, Charles A. Schreiber, and Donald A. Redelmeier, writing in *Psychological Science* in November 1993, describe Varey and Kahneman 1992. Subjects saw a series of discomfort ratings, on a scale from 0 to 10, said to come from another person every 5 minutes during something unpleasant, for example loud drilling. An unweighted combination of the peak and the end accounted for 94 percent of the variance. Adding duration raised R squared by a further 3 percent. The series 2-5-8-4, a 20-minute episode ending at 4, was judged much less aversive than 2-5-8. The only added piece was 5 minutes that ended more gently. Fredrickson and Kahneman, in the film paper, call that 94 percent a recent reanalysis of the published data. The 1992 paper was not re-read. Those profiles were marks on a page, not pain the rater underwent. [Affect Heuristic](/effects/affect-heuristic) is a judgment steered by a feeling about the object. These profiles asked for a global score of a written series.

## Key Points

Four cuts keep a noticed length, a raw correlation, a partial correlation, and a delayed ranking from inheriting one another. The films, the dial, and the later rank each have their own number.

<Steps>
  <Step title="People can report the length and still not use it">
    Duration Neglect is not a failure to tell a long episode from a short one. In Study 1, the 32 viewers estimated the long versions, on average, as twice the short ones. The clips were built to be about three times as long. Pleasant versions averaged 36.5 seconds and 110.5 seconds. Aversive versions averaged 28.8 seconds and 84.2 seconds. The estimate was a difference, not the true ratio, and the global rating still gave length little independent weight. In the cold-water session, the relative duration was judged correctly on average. The choice still favored the longer trial.
  </Step>

  <Step title="Pleasant clips did not gain a better score by running longer">
    Study 1 used 32 Berkeley students, 16 women and 16 men, ages 16 to 35, median 22. Each was paid 10 dollars for about 90 minutes. Each person saw short and long versions of plotless clips and, right after each clip, marked a 100-millimeter line for overall pleasure or displeasure. Lengthening the pleasant clips did not reliably raise that line. The within-person comparison of each viewer's three short clips with that viewer's three long clips gives t = 0.87, and a mean biserial correlation of .10. The average lines were 49.72 and 52.86. Two long versions were liked less, not more. The puppy clip moved from 33.50 to 30.38. Waves moved from 50.94 to 45.19. The prose says the long puppy film ended with mean affect on the negative side of the meter, while real-time ratings stayed positive through the waves film and almost to the end of the puppy film. Adding still-pleasant seconds did not improve the score.
  </Step>

  <Step title="On aversive clips, the raw shift follows the affect, not the clock">
    The same comparison on aversive clips did move. Means on the displeasure line were 55.18 for the short versions and 65.33 for the long ones, t = 3.08, p \< .01, mean biserial correlation .26. Moment-to-moment affect also grew more negative. A separate group of 20 viewers, shown only the long versions, said the basic content had not changed on 83 percent of aversive responses and 77 percent of pleasant responses. Among aversive responses, 72 percent said their emotion got worse, 24 percent said it did not change, and 4 percent said it grew less negative. Those are percents of responses, not a head count of 20. The test of Duration Neglect is what remains after peak and end are held fixed. For the six pleasant films each person saw, the mean partial correlation of length with the global rating was .06, t(30) = 0.72. For the six aversive films, it was −.02, t(30) = 0.18. Neither partial is marked as significant. One viewer's real-time record was unusable, so the correlations that use the meter are based on 31 people. The authors treat the partials with caution. Their preferred evidence is the pattern: small and inconsistent length effects for pleasant films, and a stronger tie of length to the live negative affect than to the global rating for aversive films.
  </Step>

  <Step title="A delayed rank made the length link smaller, and only marginally so">
    Study 2 used 96 Berkeley students, 48 women and 48 men, ages 18 to 57, median 21, paid 10 dollars for about an hour. They did not move the meter. At the end they ranked clips for a tape other people would watch, pleasant clips to maximize pleasantness and aversive clips to minimize unpleasantness. Ranks ran from 1 to 8. On-line viewers were told the ranking in advance. Delayed viewers were not. For aversive clips the mean biserial correlation of length with the rank was .24 on-line and .12 delayed. The paper says that drop reached only marginal significance, t(94) = 1.90, p \< .10. Pleasant correlations were .06 on-line and .11 delayed, neither a reliable duration effect. The authors say the delayed rank is closer to ordinary memory, because people do not usually watch an episode in order to score it later. The marginal test is not a second, sharper result. It is a smaller correlation that they do not call solid at the 5 percent line.
  </Step>
</Steps>

## Applications

Use the partial, the choice, and the return rate on the question that produced each one. A longer episode is not automatically a worse memory, and a gentler ending is not automatically a higher chance of coming back.

<CardGroup cols={2}>
  <Card title="Write the clock beside the score" icon="clock">
    In the film room, long clips were estimated at about twice the short ones, and the partial correlation of length, after peak and end, was .06 or −.02. A meeting note that keeps only "how it felt" has dropped the clock the viewers could still report. Ask for the length in the same sitting as the global line.
  </Card>

  <Card title="Read a preference for the longer trial against the ending" icon="hand">
    Twenty-two of 32 men chose the 90-second cold trial. Seventeen of the 21 whose discomfort fell by 2 points or more chose it. The 11 whose discomfort fell by 1 point or less, or rose, split 6 to 5 toward the short trial. A team that cites 69 percent without that split has erased the ending the authors use to explain the choice.
  </Card>

  <Card title="Keep two colonoscopy tallies apart" icon="stethoscope">
    In the 2003 trial, leaving the scope in the rectum for up to 3 minutes lowered the final 3-minute pain from 2.5 to 1.7 and the later discomfort mark from 4.9 to 4.4. The chance of a later colonoscopy was 53 percent against 48 percent, which was not significant. An adjusted odds ratio of 1.41 is a different test. Do not quote one and drop the other.
  </Card>

  <Card title="Do not import a forecast of future minutes" icon="hourglass">
    [Planning Fallacy](/effects/planning-fallacy) is a too-optimistic forecast of how long a task will take. Duration Neglect is a rating after the episode. A student who budgets too little time for an assignment, and a patient who later scores a long procedure by its last minutes, are not one error. Ask whether the clock is still in the future or already in the memory.
  </Card>
</CardGroup>

## Case Study

In November 1993, Kahneman, Fredrickson, Schreiber, and Redelmeier asked whether adding a gentler ending could make people choose more pain. The journal is *Psychological Science*, volume 4, number 6, pages 401–405. The author line places Kahneman and Schreiber at the University of California, Fredrickson at Duke, and Redelmeier at the University of Toronto. The correspondence address is Kahneman at the Woodrow Wilson School, Princeton University.

Thirty-two male University of California students took part. Ages ran from 19 to 39, median 22.5. Each was paid 10 dollars for a 1-hour session. They were screened for health problems and for drugs, including tobacco. Three of the 32 replaced other men whose data were discarded: one for a temperature-control failure, and two because they did not give the same preference on two choice questions. The 22 who chose the long trial are inside this replaced group of 32. They are not a count of everyone who sat down.

Each man expected three cold-water trials. The cover story was a study of lateral differences. A female experimenter ran the session. The tub held 7 liters at a mean of 14.1 °C. The short trial was 60 seconds at that cold. The long trial was 60 seconds at the same cold, then 30 seconds while the water rose by about 1.1 °C, to a mean of 15.2 °C. The abstract rounds the temperatures to 14 °C and 15 °C. The results use 14.1 and 15.2. The pump change was not announced by a sound the man could use as a cue. Order and dominant hand were counterbalanced. Before each trial both hands rested in room-temperature water for 2 minutes. Between trials he waited 7 minutes with a personality inventory. The inventory is not named.

A discomfort meter ran from 0 to 14. At 60 seconds the means were 8.44 on the short trial and 8.34 on the long trial. Over the last 30 seconds of the long trial the mean drop was 2.65, t(31) = 6.80, p \< .01. Eleven of the 32 showed a drop of 1 point or less, and two of those eleven reported a rise. The other 21 showed a drop of 2 points or more. The paper says the ending was still distinctly unpleasant for that majority.

The choice came before a third trial that was never run. One question asked which trial he would repeat tomorrow if paid. The next asked which he chose for today's third trial. The two men who disagreed with themselves were the replacements, not members of the 32. Twenty-two of the 32, or 69 percent, chose the long trial, z = 2.15, p \< .05 by a sign test. The authors note that this share would be zero if the men had been minimizing time in the water. The choice did not depend on which trial came first or on which hand was used. Preference for the long trial correlated with the size of the late drop, biserial r = .38, p \< .05. Seventeen of the 21 with a drop of 2 or more, or 81 percent, chose the long trial. The 11 with little or no drop split 6 to 5 in favor of the short trial. The paper does not print the complementary head counts. The 17 and the 5 from that split are the two groups that land on the long trial inside the 22.

After the choice, means on scales from −5 to +5 said the long trial caused less overall discomfort, M = −0.91, t(31) = 2.12, p \< .05, and was less tough to cope with, M = −1.12, t(31) = 2.90, p \< .01. The paper calls those judgments simply wrong, because the long trial contained the short trial and then more. The coldest-moment comparison was M = −0.91, t(31) = 1.90, and that test is not given a p value. Duration was judged in the correct direction on average, M = 1.09, t(31) = 3.27, p \< .01. Six men said the long trial was the shorter one. Nine reported no difference. The paper does not print how many of the remaining men called the long trial longer. Choice correlated .80 with the overall-discomfort comparison and .16 with the duration comparison.

A replication without the running meter is one sentence. Of 37 participants, 24 chose the long trial. No test is printed on that sentence. It is not a second 69 percent.

The lesson is a choice of 30 extra seconds of milder cold. The 69 percent is 22 of these 32 men, after two inconsistent men and one failed temperature setting were replaced. It is not a rate for a clinic, a class, or a household.

## Boundaries and Failure Modes

Duration Neglect is the wrong summary when a raw correlation is treated as no effect at all, when 22 of 32 is treated as a universal preference for more pain, or when a colonoscopy odds ratio is treated as the unadjusted return rate.

The first boundary is the film pattern itself. Pleasant length did not reliably change the immediate global line, and two longer pleasant clips were liked less. Aversive length did change that line, t = 3.08, because the affect escalated. The partials near zero are the neglect result. A zero-order correlation over the six pleasant films was .13, t(30) = 1.61, unmarked. Over the six aversive films it was .25, t(30) = 3.44, marked with one star. That .25 is not the biserial .26, and the .13 is not the biserial .10. They are two ways of pairing length with the rating. The authors also say length might matter when an episode is very much longer or shorter than expected. These clips were seconds to about two minutes. They warn against treating the result as a finding about days, months, prison terms, or years of illness. No such series is in the paper.

The second boundary is the clinical count, and it changes sample by sample. The film paper's discussion cites a colonoscopy series as Redelmeier and Kahneman 1992: 101 patients, durations from 4 to 69 minutes, with peak and the last few minutes predicting the retrospective rating, and duration not a significant predictor for the patient, the physician, or the nurse. The reference list instead prints a 1993 manuscript submitted for publication. This page does not merge those two years. The 1996 paper in *Pain*, volume 66, pages 3–8, was not re-read past the abstract. That abstract gives colonoscopy n = 154 and lithotripsy n = 133, correlations of total pain with the peak and with the last 3 minutes, both P \< 0.005, and says lengthy procedures were not remembered as particularly aversive. Those are not the 101, and not the 4-to-69 range.

The 2003 randomized trial is a third sample. Redelmeier, Joel Katz, and Kahneman, *Pain*, volume 104, pages 187–194, July 2003, DOI 10.1016/S0304-3959(03)00003-4. Received 3 September 2002, accepted 26 December 2002. Of 733 outpatients approached in Toronto, 682 agreed. Conventional care was 345. Modified care was 337. At the end, the colonoscope rested in the rectum for up to 3 minutes, with no suction, no inflation, and no added anaesthetic. Staff were not told the assignment until the scope reached the cecum, or until further insertion was judged not feasible. Patients were not told which group they were in. Final 3-minute pain was 1.7 against 2.5, P \< 0.001. Overall duration was 27.6 minutes against 26.8, P > 0.20. The instruction was up to 3 minutes. The measured mean gap is about 1 minute and is not significant. Later discomfort on a 10-centimeter line was 4.4 against 4.9, P = 0.006. Rank among eight unpleasant events, with 8 the worst, was 4.1 against 4.6, P = 0.002. Duration correlated .10 and .09 with those two memories. Average pain during the procedure correlated .01 with duration, P > 0.20.

Return for another colonoscopy, median follow-up 5.3 years, averaged 50.4 percent in the abstract. The body says 53 percent against 48 percent, P > 0.20. Logistic regression that controlled for prior colonoscopy, the indication, and abnormal findings gave an odds ratio of 1.41, P = 0.038. The authors call the unadjusted gap no large general effect. Among patients they classed as low propensity to return, the rates were 43 percent against 32 percent, P = 0.023, and they say the intervention yields one more return for every nine such patients treated. They call the return effect negligible in the high-propensity group. [Outcome Bias](/effects/outcome-bias) is a higher rating of the same decision once the judge knows the result. A later bill for a repeat colonoscopy is not that rating, and 1.41 is not 22 of 32.

The third boundary is the common misuse. Duration Neglect is not the claim that people cannot tell time. The film estimates and the cold-water duration mean say they often could. It is not a policy of ignoring time in sentencing, wages, or schedules. The authors say duration is used when attention is drawn to it, and that the miss is attentional. It is also not a count of lives, dollars, or people left out of a judgment. Those extensions are a different question. The cold-water paper says the peak-and-end pattern does not exhaust the predictors. A fast improvement, a hope that is later disappointed, or a memory dominated by the beginning can matter. Those cases were not the film rows or the 32 men. The 69 percent, the partials near zero, and the odds ratio 1.41 do not travel as one workplace rate.

## Common Misconceptions

Three retellings turn a partial correlation into "length never changes the score," a correct duration judgment into blindness, or an adjusted odds ratio into proof that a longer procedure brings patients back.

<AccordionGroup>
  <Accordion title="A longer bad episode is remembered as just as bad, so length does nothing">
    On the aversive films the immediate displeasure line did rise, from 55.18 to 65.33, t = 3.08, p \< .01. Discomfort on the meter rose as well. Duration Neglect is the further claim that this rise is carried by the changing affect. The partial correlation of length, after peak and end, was −.02, t(30) = 0.18. Calling the raw means "no difference" drops the result the authors actually printed.
  </Accordion>

  <Accordion title="People do not know which episode lasted longer">
    Film viewers estimated the long clips, on average, as twice the short ones. Cold-water judgments of relative duration averaged 1.09 on the comparison scale, t(31) = 3.27, p \< .01. Six men reversed the trials, and nine reported no difference. The choice correlation with that duration comparison was .16, against .80 with judged overall discomfort. The miss is the weight in the choice, not a blank where the clock should be.
  </Accordion>

  <Accordion title="Adding a gentler ending makes patients return">
    In the 2003 trial the gentler ending lowered the final pain and the later discomfort mark. The unadjusted return rates were 53 percent and 48 percent, P > 0.20. The odds ratio 1.41, P = 0.038, appears after controls for prior colonoscopy, indication, and abnormal findings. A low-propensity subgroup, 43 percent against 32 percent, is not the rate for all 682. None of these figures is the 69 percent from the cold-water room.
  </Accordion>
</AccordionGroup>

## Related Concepts

Duration Neglect sits next to a rule about which moments carry the weight, a forecast about future minutes, and a rating that moves after the result is known. Each of those pages has its own count.

<CardGroup cols={3}>
  <Card title="Peak-End Rule" icon="chart-line" href="/effects/peak-end-rule">
    The worst or best moment and the final moment can dominate the memory of an episode. Duration Neglect is the small separate weight of length, not another name for those two moments.
  </Card>

  <Card title="Focusing Illusion" icon="magnifying-glass" href="/effects/focusing-illusion">
    Attention stays on one striking part of a larger scene. Duration Neglect is that kind of miss only for the length of one affective episode already under way.
  </Card>

  <Card title="Planning Fallacy" icon="hourglass" href="/effects/planning-fallacy">
    A plan underestimates how long a future task will take. Duration Neglect starts after the episode, when a length the person can report still drops out of the global rating.
  </Card>

  <Card title="Outcome Bias" icon="scale-balanced" href="/effects/outcome-bias">
    The same decision is rated higher once the judge knows it turned out well. A cold-water choice is a preference between two past trials, not a grade of a decision after its result arrives.
  </Card>

  <Card title="Affect Heuristic" icon="heart" href="/effects/affect-heuristic">
    A feeling about an object steers the judgment of it. The film meter and the cold-water dial recorded affect over time. The global line is the later score, not that running feeling itself.
  </Card>
</CardGroup>

## One-Line Takeaway

<Tip>
  When you score a past episode, write down how long it lasted before you trust the feeling. The length can be known and still almost absent from the score.
</Tip>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.