Abstract
Previous studies regarded attention as the “gatekeeper” of visual working memory (VWM), with attended information being selected into VWM while unattended information being filtered. However, in most previous studies the attended information was also the to-be-remembered one, where attention and memory requirement was intertwined with each other. Here, we argued attention and VWM could be dissociated when requiring participants to attend to and use certain information to perform a task while not requiring them to remember it. Such information was termed as key feature, and we directly test whether the key feature has ever been encoded into VWM in Experiments 1-3 through a variety of paradigms. We found that the key feature could not produce a working memory-driven attentional bias effect (Experiment 1), consumed no VWM capacity (Experiment 2), and produced no CDA component (Experiment 3), indicating that the key feature, despite being fully attended and used, was not encoded into VWM. These results suggested a highly selective VWM gate enabling selection even among attended information. In Experiments 4-7, we further explored how the VWM gate worked to achieve selection among attended information. Specifically, when the key feature was attended, we forced participants to open the VWM gate by asking them to simultaneously memorize another item. Interestingly, we found that a key color would enter VWM when the memorized item was also a color (Experiments 4 and 5), but not when it was a shape (Experiment 6). Experiment 7 found that a key shape would enter VWM when the memorized item was also a shape. These results suggested a feature-specific VWM gate for attended information. That is, we could selectively open the color gate for one attended item while close the shape gate for another attended item. However, once the color gate is open, all attended colors would enter VWM.