THE REPRESENTATIONS OF VISUAL WORKING MEMORY: WHERE, WHAT, AND HOW

Visual working memory allows humans to remember things that they saw a few moments ago and that memory can last for a short period of time and it is an essential function for a lot of cognitive processes. This dissertation focuses on three basic questions regarding visual working memory. Specifically, where is visual working memory (not) maintained in the human brain? In what format is it represented? And how information exerts its influence on the memory capacity? Firstly, I investigated visual working memory performances under visual interferences, across a variety of stimuli categories. Once consolidated, visual working memories are intact even in face of drastic disruptions, in huge contrast with performances of iconic memory, which is known to rely on sensory regions. The contrast suggests that visual working memory is unlikely to be maintained in the early visual areas as the sensory recruitment theory of visual working memory claims. Instead, higher-order brain regions are involved to safely store robust memory contents. To further understand the representations of visual working memory, I examined the formats of visual working memory and the imagery-versus-propositional debate in specific. Comparing across conditions, I found visual working memory contents consisted of analog constituents which are bound together. Single features like color and orientation are represented in a continuous fashion whereas the bindings between features are brokable to an extent like discrete propositions. This leads to a follow-up question: whether representations are created equal in visual working memory? It is known that items in the same feature domain are not represented equally in visual working memory, this dissertation looks at the binding components, are they created equal? Given the same experiment settings where the only difference is the spatial layouts of bindings between features, participants committed more swap errors when stimuli are vertically stacked than horizontally placed, suggesting a difference in binding strength in memory representations. Lastly, this dissertation explored the relationship between information and the limited capacity of visual working memory representation, and the key process that mediates the relationship. Complexity limits the memory capacity, and this effect is established via the consolidation process