Our Research

In the auditory system, function often directly follows form. Segregation of cell types in the organ of Corti and cochlear nucleus has revealed fundamental lessons about the biology of hearing. However, no such understanding exists at the level of the auditory cortex. By disambiguating the roles of distinct classes of projection neurons, our research program aims to understand the principles of cortical organization, and how such principles can give rise to complex behaviors.

Our Approach

We observe neural dynamics in a number of ways. We monitor large-scale network activity of genetically identified cell-types in the auditory cortex during using two-photon calcium imaging, and we then perturb those networks using optogenetics and spatial light modulation. This allows us to characterize large neural populations and to understand how local networks of different cell-types interact with themselves and each other.
We monitor and perturb neural activity of genetically identified cell-types within the auditory cortex and various downstream projection targets using chronic electrophysiology combined with opto- and chemo-genetics. This allows us to understand how information in the auditory cortex is used by different downstream areas.
Both of these techniques for observing and perturbing neural circuitry are carried out in behaving mice, to understand cortical circuit contributions to auditory-guided behavior.

We use viral tracing strategies to understand the downstream targets of genetically identified cell-types in the auditory cortex, and the specific inputs that these cell-types receive.

We study auditory-guided behaviors using head-fixed mice that navigate a virtual reality during the presentation of different auditory cues. Using virtual reality for such a task allows for voluntary trial initiation, ensures active listening, and allows for rapid modification of the environment in a way that would be impossible with conventional behavioral training chambers.

Advances in experimental techniques allow for the simultaneous observation of thousands of neurons. Such high-dimensional data sets can be difficult to analyze and interpret. We use techniques from machine learning to build statistical models that can characterize neural population responses and extract structure from high-dimensional neural data.