This project will investigate how listeners are able to comprehend speech in challenging acoustical conditions like cocktail parties. Speech-recognition technology dramatically underperforms human hearing in such conditions, which has been a barrier to improving voice-based computer interfaces and prosthetic speech-to-text assistive devices. This research will address this gap by examining a perceptual illusion called auditory restoration, in which the brain fills in words and other sounds that have been interrupted by brief, loud noises. This phenomenon implies that the brain constructs robust internal models of speech sounds, and that it uses them to determine what is being said from limited or degraded information. To learn how these internal models work, this study examines perceptual restoration in zebra finches, a species of songbird. Zebra finches employ complex vocalizations to identify themselves to each other, and need to be able to understand who is singing in noisy social settings. The project combines methods of measuring behavior and brain activity to probe the neural circuits underlying auditory restoration in finches. By identifying these circuits and how they are wired up, this research will give insight into fundamental perceptual processes that allow humans to communicate through speech, and more broadly, that enable the brain to construct coherent experiences from sensory information that is often incomplete and unreliable. A better understanding of the neural mechanisms of auditory perception also promises to enhance the performance of speech-recognition algorithms to facilitate communication. The educational component of the proposal addresses a growing need in the field for computational proficiency at earlier stages in training. The investigator plans to implement a curriculum to encourage and support early exposure to computational and systems neuroscience, serve as a recruitment tool for systems neuroscience labs, and contribute to the university's goals of increasing data literacy and skills in data science.

Auditory restoration has been widely observed in a broad range of mammalian and avian species. It is likely to contribute to humans’ remarkable ability to reliably decode phonemes from speech in extremely noisy conditions. The research will advance the understanding of this phenomenon by bringing a well-characterized animal model of auditory processing to investigate a well-established human psychophysical paradigm. The approach combines in vivo extracellular and intracellular electrophysiology with operant behavior to examine how bottom-up and top-down processes interact within the auditory cortex to support auditory restoration in zebra finches. Aim 1 examines how auditory experience affects behavioral auditory restoration. Aim 2 determines where the neural correlates of top-down and bottom-up restoration emerge in the auditory processing hierarchy. Aim 3 begins to dissect the mechanism by investigating how specific cell types and their intrinsic neural dynamics contribute to tolerance for noise. The educational objectives include the development of a curriculum to train students in data science and best software practices for systems neuroscience and will impact the broader community through freely distributed, interactive instructional materials and training the next generation of data scientists in neuroscience.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

National Science Foundation (NSF)
Division of Integrative Organismal Systems (IOS)
Application #
Program Officer
Sridhar Raghavachari
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Virginia
United States
Zip Code