This experiment investigated the effect of the short-term retention of duration on temporal discrimination in 5- and 8-year-olds, as well as in adults, by using an episodic temporal generalization task. In each age group, the participants' task was to compare two successive durations (a standard and a comparison duration) separated by a retention interval of 500ms, 5s, or 10s, with the order of presentation of these two durations being counterbalanced. The results revealed a shortening effect for the first presented stimulus in all of the age groups, although this was greater in the younger children, thereby indicating the presence of a negative time-order error. Furthermore, introducing a retention delay between the two durations did not produce a shortening effect but instead flattened the generalization gradient, especially in the younger children. However, this flattening of the generalization gradient with the retention delay was more marked between 500ms and 5s than between 5s and 10s. Thus, retaining the first duration in short-term memory during a task requiring the comparison of two successive durations reduced temporal discrimination accuracy and did so to a greater extent in the younger children.