Duration Neglect: Why How Long It Lasted Barely Counts

Seconds are what an experience is made of and not what it is remembered by. Asked how bad a past episode was, people answer with something close to the average of its worst moment and its last moment. The number of seconds it ran, which would dominate any total of the discomfort, contributes almost nothing. The peak-end rule names the moments that do the work; this page is about the quantity that goes missing.

The clearest sign that a total is not being computed is that the judgments can be reversed by adding discomfort. An episode rated 2, 5, 8 at successive intervals is judged worse than an episode rated 2, 5, 8, 4, although the second contains the first and then more. Nothing was subtracted. A better final moment changed the summary, and the extra minutes did not enter it.

The mechanism behind this is not settled, but the shape of the answer is agreed. What is kept is a few moments that stand in for the episode, and no single moment can stand in for its length: duration is not a property of any instant, it only exists once the instants are added up, and nothing in memory adds them up unasked. The consequence is practical rather than philosophical. Decisions about whether to repeat an experience are made from the summary, so they can go against the total: a longer course of discomfort with a gentle ending can be chosen over a shorter one that stopped at its worst.

Duration is neglected rather than ignored, and the page would oversell by dropping that. Asked directly how long something took, people answer accurately, so this is not an inability to count. And length returns once it is big enough to become the story: Kemp, Burt and Furneaux applied the peak-end account to extended autobiographical events such as vacations in 2008 and reported that it was not an outstandingly good predictor there. The finding is about the weight duration carries inside a global evaluation, which is small, not about whether it can be reported on request.

Barbara Fredrickson and Daniel Kahneman published "Duration neglect in retrospective evaluations of affective episodes" in the Journal of Personality and Social Psychology in 1993, and the term comes from that paper. Thirty-two participants watched 12 test clips, short and long versions counterbalanced across participants, reported their feelings continuously and then gave a global evaluation of each clip. The unpleasant clips averaged 29 seconds short against 84 long. The long versions were rated worse, but the whole of that difference was accounted for by discomfort escalating while the film ran. With the moment-by-moment record held constant, the extra seconds added nothing, which is what the term duration neglect names. Participants could tell the versions apart: asked how long each clip ran, they put the long ones at about twice the short ones.

The number on duration's share comes from the same 1993 paper, in a reanalysis of data Carol Varey and Kahneman had published a year earlier in the Journal of Behavioral Decision Making. Their participants judged other people's episodes of discomfort from schematic reports: a rating every five minutes through an episode of 15 to 35 minutes. In those published judgments the peak rating and the final rating together, with no weighting at all, accounted for 94 percent of the variance, and duration, although significant, added 3 percentage points. That 3 against 94 is the figure this page rests on: the length of every episode was printed in front of the reader, and it did almost no work. The demo above holds the peak and the ending fixed and varies only the length, which is the comparison that leaves duration nowhere to hide.

The best-known demonstration came with real discomfort. Kahneman, Fredrickson, Charles Schreiber and Donald Redelmeier had 32 participants put one hand in water at about 14 degrees Celsius for 60 seconds, and the other hand in the same water for 60 seconds followed by 30 further seconds during which the temperature was raised by about one degree, still unpleasant but less so. Twenty-two of the 32, or 69 percent, said they would rather repeat the long trial. The long trial was the short one with 30 further seconds of cold added to the end, so that preference reverses the total: more discomfort, chosen freely, because of where it stopped.

The hospital version of the result, in 1996, is the one that shows how wide the neglected range can get. Redelmeier and Kahneman recorded pain minute by minute through 154 colonoscopies, and those procedures ran from 4 minutes to 67. The longest took more than fifteen times as long as the shortest, and that difference left no mark on how bad patients afterward said the procedure had been: what their judgments followed was the worst minute and the last few.

The pattern appears wherever an experience is summarized afterward and the summary is used to decide. The rows below are illustrations, not findings from the studies cited on this page.

Duration is not hard to measure, so the counter is to measure it rather than to recall it.

• Keep a record in real time. A note at intervals during a long task, a shift or a treatment preserves the thing the summary leaves out.
• When a decision is about repeating something, write the total down: how long, how often, how bad on average. Compare totals with totals rather than impressions with impressions.
• Ask about the parts separately. How long did it take, how bad did it get, how did it finish. Three questions resist the summary that one question invites.
• Watch for endings doing the work. An experience arranged to finish well will be remembered well, which is useful to know when you are the one arranging it and worth noticing when someone else is.
• For pleasant experiences, spend on the parts that will be lived rather than the parts meant to be remembered, and expect that doubling the length will not double the memory.

Three nearby effects concern the gap between an experience and its memory. They differ in what is distorted.

• It is not the peak-end rule, which is the other half of the same finding. The rule says the worst moment and the final moment predict the judgment; duration neglect says the length of the episode does not.
• It is not rosy retrospection, the tendency to recall an episode as better than it was rated at the time. That shifts the tone of a memory; duration neglect leaves out a quantity.
• It is not the impact bias, the tendency to overestimate how strongly and how long a future event will affect you. That is a prediction about feelings to come; duration neglect is an evaluation of an episode already finished.