Abstract
This study investigates the release of information from visual short-term memory (VSTM) when it is no longer needed. Previous studies have found inconsistent results, with some studies showing effective release of irrelevant information (Olivers, Meijer & Theeuwes, 2006), and other studies showing proactive interference effects from previously remembered items (Jonides & Nee, 2006). Here we investigate whether visual information retained on trial N-1 continues to interfere with performance on trial N, whether this interference is spatially specific, and whether it relies on active memory from trial N-1. Using change detection tasks of colors or shapes, we show that participants tended to falsely classify a changed item as “no-change” if the item matched one of the memory items on trial N-1, suggesting that memory for trial N-1 was not effectively released. The interference is spatially specific: memory for trial N-1 items interfered more if the test item match the feature value and the location of a trial N-1 item, than if it did not matched the location. Yet, the interference is not retinal-based and is seen when the display on trial N-1 was expanded or contracted compared with trial N. A four-alternative-forced-choice task revealed that intrusion from trial N-1's item was similar in magnitude to intrusion from trial N's adjacent items, suggesting that temporal interference is comparable to spatial interference. Finally, interference from items on N-1 is a direct consequence of retaining information in VSTM. If participants passively viewed items on trial N-1, no intrusion was found. We conclude that humans are not efficient at releasing unwanted information from memory, even though intuitively its better to delete previously remembered items to make room for new input.