Problem & approach
Problem. Flexible movement requires control at multiple levels of abstraction: from high-level planning to low-level feedback. The brain appears to achieve this through a distributed hierarchy, with different areas operating on distinct timescales and representational granularities. However, we still lack a mechanistic understanding of how this hierarchy is formed, represented, and updated through learning.
Approach. I study hierarchical control as a property of neural systems, focusing on how different brain regions contribute to separate levels of control and how these levels interact and adapt over time. By analyzing large-scale population recordings during complex motor tasks, I aim to uncover the neural mechanisms that allow hierarchical organization to emerge and to identify how learning signals and credit assignment are distributed across this hierarchy.
Highlights
- Mapping how different brain regions correspond to distinct levels of the motor hierarchy.
- Identifying representational transformations between hierarchical levels.
- Examining how inter-area communication coordinates planning, execution, and feedback.
- Understanding how the brain assigns credit across levels when movement errors occur.
Questions I’m exploring
1. Mapping the hierarchy across brain regions
- Which brain areas correspond to distinct control levels (e.g., premotor → planning, M1 → execution, cerebellum → feedback)?
- Are these levels organized along a gradient of temporal abstraction or representational detail?
- How do cortical, subcortical, and spinal circuits interact to coordinate action?
- Does learning reshape this division of labor, or is the hierarchy anatomically fixed but functionally flexible?
2. Representational format and transformation
- What is the format of neural representations across control levels?
Are higher-level areas encoding intentions or policies while lower areas encode feedback dynamics? - Are low-level manifolds nested within high-level population spaces?
- How are representations transformed between regions?
- How does context or experience reconfigure these mappings during learning?
3. Coordination and communication
- How do regions operating at different control levels communicate and synchronize during movement?
- Are there distinct pathways for top-down commands versus bottom-up feedback?
- What neural signatures mark switching or delegation between hierarchical levels?
- Do transient dynamics or oscillations mediate inter-area coordination?
4. Learning and adaptation
- How does hierarchical organization emerge through learning?
- What changes in neural activity accompany the formation or refinement of control hierarchies?
- How do different regions co-adapt when task dynamics change?
- What principles govern stability vs. plasticity when learning spans multiple control levels?
5. Credit assignment across the hierarchy
- When a motor error occurs, how does the brain decide where to assign blame—to the plan, sequence, or execution?
- Are there multiple parallel error signals across levels, or a single global signal interpreted locally?
- How do feedback and learning signals propagate through cortical–subcortical–cerebellar loops?
- How is stability maintained when credit assignment must span different timescales and levels of control?
Results / status
I’m currently developing analyses to identify hierarchical relationships in neural population dynamics, and testing whether learning reshapes how different regions contribute to planning and feedback.
Future work will quantify cross-area communication and error propagation during adaptive control.
Highlights
- Neural hierarchies: uncovering how distinct brain regions implement layered control of movement.
- Learning and adaptation: understanding how hierarchical structure forms and changes with experience.
- Credit assignment: revealing how error signals and learning rules are distributed across the hierarchy.
Related publications
See all publications tagged with this project:
View related publications →