It is a common time management problem that people underestimate the duration of tasks, which has been termed the “planning fallacy.” To overcome this, it has been suggested that people should be informed about how long they previously worked on the same task. This study, however, tests whether previous misestimation also affects the duration estimation of a novel task, even if the feedback is only self-generated. To test this, two groups of participants performed two unrelated, laboratory-based tasks in succession. Learning was manipulated by permitting only the experimental group to retrospectively estimate the duration of the first task before predicting the duration of the second task. Results showed that the experimental group underestimated the duration of the second task less than the control group, which indicates a general kind of learning from previous misestimation. The findings imply that people could be trained to carefully observe how much they misestimate task duration in order to stimulate learning. The findings are discussed in relation to the anchoring account of task duration misestimation and the memory-bias account of the planning fallacy.