Abstract
Retro-cue paradigms show that we can prioritize some items in memory at the expense of other items. The most common explanation is that memory resources are reallocated from uncued to cued items. Here we test this popular theory by employing a working memory task with unequal item rewards to create uneven distribution of resources. Retro-cues then indicated that participants could remember cued items and drop uncued items from memory. If reallocation occurs, then we should observe a larger benefit when high reward items, which consume more resources, become irrelevant. Experiment 1 found that memory was better for items with higher reward values, F(2,110)=15.00, p<.001, and for valid versus neutral cues, F(1,55)=22.60, p<.001. However, dropping a higher reward item does not benefit performance more, F(1,55)=0.19, p=.662, inconsistent with reallocation. Experiment 2 replicated this null finding on a large data set (n=100) and with a higher memory load, and further showed that the disengagement of resources from a high versus a low reward item can benefit the encoding of a second array. Participants encoded two sequentially presented arrays. Items in the first array had unequal reward. Participants were later instructed to drop either high or low reward items from this array. A benefit for dropping high reward items was found when the cue was shown BEFORE the second array t(99)=4.54, p<.001, but not AFTER the array was already encoded, t(99)=1.00, p=.321 (interaction was p=.030). Thus, there was no benefit of removing high resource items for information already encoded into memory, but there was a benefit for encoding new information. Together, these results suggest that retro-cues provide a benefit to stored information (e.g. due to protection or strengthening) that is, in contrast to the popular view, not linked to the withdrawal of resources from uncued items.