Past, present, future. On the action-oriented nature of visual working memory

For visual working memory to guide upcoming behavior, it is crucial that we prepare for the potential use of its contents ahead of time. Recent studies have demonstrated how the prospection and planning for an upcoming manual action starts early after visual encoding, and occurs alongside visual retention. In this thesis, I build upon this line of research in three ways. In chapter one, I show that action planning can occur alongside multi-item visual working memory, where multiple potential actions are prepared in parallel even when uncertain. Using a task with three memory-load conditions (one/two/four items) and a delayed reproduction report, I measured EEG signals of action planning (beta-band activity contralateral to the response hand). I observed attenuated beta activity not only for a single certain action but also when two potential future actions were required. Moreover, potential action planning in load two occurred regardless of response similarity and predicted the speed of later behavior. These findings demonstrate that multiple potential actions can be planned alongside visual working memory, reinforcing the idea that working memory is ultimately future-oriented. In chapter two, I asked whether output planning flexibly adapts to different visual-motor mappings, and whether it occurs even when an action will only potentially become relevant. I designed a sequential visual-motor working memory task in which participants memorized the tilt of one or two colored bars, with tilt-to-response mappings counterbalanced between blocks. EEG revealed that action planning reflected anticipated task demands rather than fixed mappings, and again occurred even when actions were only potentially relevant. Importantly, increased planning for one item facilitated performance for that item but came at the expense of other working-memory content. These results highlight the flexible and future-oriented nature of visual working memory, showing how anticipatory dynamics translate visual representations into upcoming behavior. In chapter three, I turned to the selection stage at the end of the memory delay, when representations are actually turned into action. Here I asked whether the slowing of memory-guided behavior under higher load is due to slower access to sensory representations or to reduced readiness to act upon them. Participants memorized the orientation of two or four bars, with one later cued for reproduction, while EEG tracked the timing of visual (cued location) and motor (manual response) selection. Results showed that slower behavior under higher load was linked to delayed motor selection, not slower visual access. Variability in response times was also explained by motor, but not visual, selection dynamics. This indicates that action readiness, rather than sensory access, is the key bottleneck in memory-guided behavior. Together, these chapters highlight the anticipatory and dynamic nature of visual working memory. They show how working memory is not a passive store of visual information, but actively prepares for potential future actions, flexibly adapting to task demands and planning even multiple possible courses of action. At the same time, the work reveals limits: while several potential actions can be prepared, the readiness to act may constrain memory-guided behavior more strongly than the accessibility of visual representations themselves. Overall, this thesis provides new insight into the neural mechanisms by which visual working memory supports action planning, offering a comprehensive perspective on how the brain uses visual representations to shape adaptive behavior.