Halassa Lab

Project 3

Frontal-hippocampal circuits for hierarchical planning:

Flexible behavior in complex environments often requires hierarchical planning, where extended action sequences are organized around intermediate subgoals. We study how hippocampal-frontal cortical circuits support this computation across spatial and abstract domains. Our central hypothesis is that the hippocampus facilitates the discovery of useful subgoals from experience while frontal cortical circuits use those subgoals to guide hierarchical planning.

To test this idea, we combine mathematical theory, recurrent neural network models, and neural recordings during a hierarchical planning task. We have developed a multi-area circuit theory in which hippocampal replay supports subgoal discovery through a plasticity mechanism optimizing for a normative clustering objective. On the other hand, an orbitofrontal-anterior cingulate circuit performs hierarchical planning over these learned abstractions: orbitofrontal cortex computes subgoal values while anterior cingulate cortex selects action sequences that maximize future subgoal values. This framework generates experimentally testable predictions on how the frontal cortex and hippocampus interact during subgoal discovery, option selection, and flexible replanning. Our long-term goal is to establish a circuit-level theory of hierarchical planning that links neural dynamics to underlying computation.