top of page

What is the role of sleep in the maintenance of information encoded by networks of neurons?

Decades of research support the idea that sleep benefits the consolidation of memories and facilitates learning. However, whether this consolidation role is mediated by removing unnecessary noise or reinforcing relevant neuronal activity is unknown. The activation of spatially selective neurons (place cells) in the hippocampus of mice during sleep is proposed to represent a form of “replay” necessary for memory maintenance. Interestingly, premotor neurons in the zebra finch cortex also become active during sleep. The song circuit in zebra finches is exclusively involved in singing, and its reactivation during sleep is proposed to replay the learned song. Our work demonstrates that neuronal activity in sleeping and singing zebra finches can be recorded using high-density silicon probes. We have also implemented machine learning approaches that accurately decode the bird’s song from brain activity when the bird is awake. However, the same process applied to the bursting neuronal activity during sleep reveals a distinct repertoire of song-like sounds which do not match the bird’s own song. Decoding the acoustic properties of a “dream” has never been reported, and we propose using zebra finches to investigate sleep reactivation and its role in learning and maintaining behavior. The highly stereotypic property of the learned zebra finch song provides a unique opportunity to interface brain and behavior and identify sleep's role in maintaining a stable behavior.

What is the cause and purpose of spontaneous deviation in learned vocal behaviors?

Writing or speaking are complex motor behaviors learned over years of practice. However, even the best writers or speakers can make mistakes. For example, filler sounds (uh, um) are one type of speech disfluency observed in every language. Complex learned motor behaviors, such as speech, require accurate communication between distributed cortical networks of neurons. We hypothesize that these brief lapses in speech patterns arise due to abnormal information integration in cortical circuits. We further propose that sub-cortical neurons generate filler sounds as an error-correction mechanism to enhance cortical communication. In order words, filler sounds represent a mechanism in which the brain tries to fix itself. We will test these hypotheses by using songbirds (zebra finches) as a model of vocal learning. These songbirds are master singers, but they can also make mistakes. Ultimately, we seek to identify how networks of neurons controlling singing behavior communicate and how they adapt to correct errors during singing.


How do social demands affect computation by distributed networks of neurons?

A large fraction of our daily routine involves transitioning from one task to another, such as walking between buildings or talking to different people. This fundamental cognitive flexibility is impaired in individuals with autism spectrum disorder (ASD), who often prefer a predictable environment and have difficulties transitioning between tasks. Dysfunction of the midbrain dopaminergic system can lead to social deficits and repetitive behaviors. However, the neuronal mechanism controlling changes in brain states during task switching and the role of dopamine is not well understood. Juvenile male zebra finches naturally alternate between a babbling and unstructured song in isolation (undirected song) to a structured and adult-like song when socially interacting with a female (directed song). However, the precise transition in neuronal computation required for such a drastic change in behavior and the role of visual feedback remains unknown. Using multi-shank and bilaterally implanted Neuropixel probes, we propose to record neuronal activity in the motor cortex, auditory cortex, visual cortex, and basal ganglia of juvenile male finches. Recordings from these brain areas will reveal the changes in neuronal activity associated with the transition between directed and undirected singing. These studies' central hypothesis is that changes in coordinated neuronal computation by distributed cortical and striatal neuronal networks mediate cognitive flexibility. These experiments will provide a mesoscale view of brain activity and fundamental insights into dopamine's role in controlling the transition between two behaviors.


How do distributed neuronal networks reorganize to recover from ischemic lesions?

Cerebrovascular diseases account for 85 % of all deaths due to neurological disorders. Of particular interest are ischemic lesions, which cause a range of cognitive and motor impairments. Speech is a complex motor behavior often impaired following an ischemic stroke. However, current rodent and primate animal stroke models cannot produce learned vocalizations. Zebra finches are a well-established animal model of vocal learning, and the neuronal networks controlling this vocal behavior have been extensively studied. However, how neuronal activity and motor behavior in songbirds are affected by ischemic lesions remains completely unexplored. In preliminary work, I demonstrated that spatially controlled ischemic lesions lead to reversible degradation and behavior recovery in adult zebra finches. We can record neuronal activity in these brain areas before and after an ischemic lesion using imaging and electrophysiological approaches. In the future, we will combine these technical advances with spatiotemporally controlled ischemic lesions to identify the mechanism regulating neuronal activity, behavior degradation, and recovery following an ischemic lesion.


How do distributed neuronal networks error-correct information relayed between brain areas?

Why humans use filler sounds is largely unknown, but it is interesting that people with autism routinely engage in self-stimulatory behaviors such as “humming”. We propose that the innate sounds produced by songbirds when they make a mistake in the learned part of the song serve as a self-stimulatory behavior. In turn, these self-stimulatory innate sounds activate dopaminergic circuits and can function as a soothing reward. We hypothesize that miscommunication between cortical brain areas leads to the activation of neurons in a dopaminergic nucleus that projects to HVC the cortex and resets its activity. In this project, we will determine the impact of call production on HVC neural dynamics and the relationship of dopamine with the reset of HVC neuronal activity.

bottom of page