The peak-end rule describes how people evaluate a past experience in memory: not by its total duration or its average intensity moment-to-moment, but almost entirely by how it felt at its most intense point and at its very end. This means two experiences with identical total suffering can be remembered completely differently depending purely on how they concluded — and, more strikingly, that a longer version of an unpleasant experience can be preferred in memory over a shorter one, if the longer version ends on a slightly better note.
Kahneman frames this as evidence for two genuinely separate systems inside a person: the "experiencing self," who lives through events moment by moment, and the "remembering self," who evaluates and stores the memory afterward — and it's the remembering self, using this distorted peak-end shortcut, that actually makes future decisions about whether to repeat an experience.
In a mild-pain lab experiment, participants twice submerged a hand in painfully cold water: once for 60 seconds, and once for 90 seconds, where the last 30 seconds involved slightly less cold (but still unpleasant) water. When later asked which trial they'd prefer to repeat, most participants chose the longer, 90-second trial — because their memory was shaped by the improved ending, not the fact that it involved 50% more total discomfort. Kahneman and physician Don Redelmeier found the same principle in a real medical setting: patients undergoing colonoscopies reported real-time pain every 60 seconds during the procedure, and their later overall pain rating was best predicted by the pain at the worst moment and at the very end — not by the total duration or cumulative pain, which had almost no effect on remembered pain at all.