Refreshing is assumed to reactivate the contents of working memory in an attention-based way, resulting in a boost of the attended representations and hence improving their subsequent memory. Here, we examined whether the refreshing-induced memory boost is a constant or a gradual, time-dependent phenomenon. If the beneficial effect of refreshing on memory performance is due to the information being selected for refreshing (i.e., selection hypothesis), a constant memory boost is expected to occur each time an item is selected for refreshing, with better memory performance for items that are selected more often. If, however, the beneficial effect of refreshing on memory performance is due to spending time in the focus of attention during refreshing (i.e., duration hypothesis), a gradual memory boost is expected, with the size of the memory boost being a direct function of how long the item has been the object of focused attention. To distinguish between these hypotheses, we instructed and guided the use of refreshing during retention through the presentation of cues, and varied the number of refreshing steps and their duration independently. The number of refreshing steps, but not their duration, had an effect on recall, in agreement with the selection hypothesis. However, some of the results were less robust than anticipated, indicating that the effect of instructed refreshing is limited to certain task parameters.