9 December 2022
Alla Karpova
HHMI Janelia Research Campus
Our work over the past seven years has laid the technical and behavioral foundation for identifying simplifying principles underlying intelligent behaviors, using rats as a model system. Recent evidence suggests that rats’ reliance on abstraction of environmental structure may share fundamental principles with hierarchical conceptual reasoning central to human cognition. Building on the intuition that animals approach complex environments by attempting to create models of the environment’s latent structure, we record and perturb ensemble activity in tasks with carefully engineered complexity. Our behavioral framework permits us to tune task complexity up and down, to build in hierarchical structured relationships and to ascertain whether an animal’s solution captures the latent structure and begins to resemble conceptual reasoning. Our early explorations within this framework of the neural dynamics in the anterior cingulate cortex (ACC) – an area implicated in keeping track of higher order abstractions, but with what previously seemed like only modestly task-related responses – have revealed surprisingly interpretable activity patterns. Using novel circuit dissection tools that we have developed with our colleagues at Janelia, we have been able to provide causal evidence for the behavioral relevance of these dynamics, and have begun to map them onto specific interacting sub-circuits within the ACC. Our findings support the notion that stronger and more organized dynamics are likely to emerge in challenging behavioral environments and suggest that it will be possible to ground even abstract cognitive computations in mechanistic insight. Going forward, we will continue to examine, in a systematic fashion, how animals' approaches to complex environments constrain neural dynamics in the frontal cortical areas, with a particular emphasis on the neural dynamics that accompany the learning and use of structured relationships and abstract concepts in behavioral tasks.