GREEN LAB

We explore how errors are detected and used for learning in the mouse cerebral cortex, a structure that expanded dramatically during human evolution, and that supports a wide range of intelligent functions -- from sensory processing to motor planning and execution. Our research directions include uncovering the roles of 1) cortical cell types, 2) neuromodulators, and 3) subcellular compartments. In addition, we compare mechanisms across sensory, cognitive and motor regions toward uncovering general principles for how the cortex learns. We have developed a range of behavioral paradigms in virtual reality to explore these questions.

Cell types

Single cell transcriptomics has uncovered dozens of molecular cell types in the cerebral cortex. What are the roles of each cell type in error signaling and learning? Building on our work on identifying a cell type that signals errors, we aim to comprehensively map the broader circuit for detecting and learning from errors. We perform in vivo two-photon calcium imaging during virtual reality behaviors designed to elicit errors, combined with enhancer viruses and spatial transcriptomics to identify cell types with new levels of precision. We also use two-photon optogenetics to identify connectivity patterns between cell types.

Neuromodulators

Cortical circuits are tuned by a handful of neuromodulators that are thought to be important for attention, learning and mood. Each neuromodulator can bind several receptors, and each receptor is expressed in a cell type–specific manner. How does each neuromodulator tune each cortical cell type, and how does this tuning impact the circuit's ability to signal and learn from errors?

Subcellular Information Processing

Cortical pyramidal neurons are composed of multiple subcellular compartments that receive different inputs and that are regulated by different inhibitory cell types. How do these compartments contribute to error signaling and learning?

General principles for learning across the cortex

The cerebral cortex is striking in that its constituent cell types are largely conserved across areas, even as different areas perform very different roles. To what extent do the mechanisms of error signaling and learning generalize across cortical regions? Are these mechanisms tailored to different functions in interesting ways?