Age-related differences in updating working memory were investigated in two experiments using a running memory task. In the first experiment, the task of the young and elderly subjects was to watch strings of four to 10 consonants and then to recall serially the four most recent items. Results revealed no age effect. A second experiment was then carried out using a memory load that was close to memory span: lists of six to 12 consonants were presented and subjects had to recall the last six items. Age interacted with list length but not with serial position. This dissociation is discussed in terms of Baddeley's (1986) model.