My primary interest lies in the neural mechanisms that translate high-level goals into concrete instructions to guide behaviors. This is implemented by studying the formation and transformation of visual working memory (WM). My research tracks the complete evolution of a visual signal, detailing how it is selected into WM (sensory codes), how it is flexibly maintained and transformed (mnemonic codes), and how it is ultimately prepared for movement (motor plans) based on the specific goal at hand. Here, I listed some research questions that I am currently pursuing.

Goal-directed WM Codes

Working memory is essential for goal-directed behaviors. WM representations are abstract and highly flexible, allowing them to be dynamically reorganized to meet changing task demands. My research centers on understanding the neural and computational processes behind this flexibility. Specifically, I address three core questions: How do high-level goals reorganize the neural codes stored in the brain? How do different brain regions interact to implement these functions? What are the underlying computational principles that define this dynamic process?

• Visual working memories are abstractions of percepts
• Putting visual working memory to work

Sensory-to-Motor Transformation

A key challenge for the brain is maintaining sensory information while simultaneously transforming it into motor codes—the format necessary for executing future actions. My research delves into the fundamental differences between these sensory and motor systems and examines how they interact and coordinate during the critical phase of motor planning. Currently, I focus on comparing these planning processes across two distinct output modalities: eyes (saccades) and hands (reaching movements).

• Effector general representation of movement goals in human frontal and parietal cortex
• Sensory codes in motor areas (ongoing)

Interactions between WM and LTM

While WM offers the brain a limited, flexible workspace for immediate action, the vast capacity of long-term memory (LTM) stores learned regularities and prior knowledge. My research explores the critical bidirectional link between these two systems. Specifically, I investigate two central questions: How do neural signals of WM predict their fates in LTM? How do learned regularities and prior knowledge stored in LTM influence the current formation of WM representations?

• Temporal expectation triggers removal of irrelevant information from working memory that leads to forgetting

Research Methods

• 🙋‍♀️ Behavioral: psychophysics, eye-tracking, motion capture system, online experiments
• 🧠 Neuroscience: fMRI, EEG
• 💻 Computational: machine learning, Bayesian modeling, neural networks