Jonathan Green, Ph.D.

Our research explores how we learn from our mistakes—a fundamental building block of intelligent behavior. As infants learning to walk, we take unsteady steps, fall, adjust, and eventually develop stable locomotion. This iterative cycle of action, error, and refinement continues throughout life, shaping our abilities to communicate, socialize, plan, and more. Although each learning scenario involves distinct sensory and motor sequences, they all share a common reliance on an error feedback loop. Our central aim is to uncover how the brain implements this core loop, toward revealing a framework underpinning intelligent behavior and its diverse expressions.

We specifically investigate how error-driven learning is mediated by the cerebral cortex—a crucial brain structure for sensory, cognitive, and motor processing. Building on our discovery of a cortical cell type that signals error-corrections, we seek to answer two fundamental questions: (1) How are errors detected? and (2) How are these errors used to guide learning?

To address these questions, we systematically study the contributions of cortical cell types, the neuromodulators influencing their activity, and the subcellular electrical and molecular events orchestrating their functions. To precisely interrogate each cell type, we have contributed to the development of viral strategies for targeting cell types in the brain. Additionally, we apply spatial transcriptomics to simultaneously map all cell types within specific cortical regions. By integrating these methods with in vivo two-photon imaging, we capture detailed subcellular neural activity and molecular dynamics across cell types in behaving mice as they engage in tasks such as virtual reality navigation and motor coordination.

Our research follows two complementary paths. First, we examine the role of the premotor cortex in using error feedback to refine and update motor plans. Second, we compare the functional roles of cell types across motor, cognitive and sensory regions, leveraging the remarkable consistency of cell types across the cortex. Through this comparative approach, we aim to identify general principles of error-based learning across the cortex, toward advancing a unified framework for cortex-based intelligence.