Variability and robustness in birdsong
Zebrafinch birdsong is an excellent model for skill learning and maintenance, as birds learn a single song through trial and error and maintain it throughout their lives. We are interested in understanding the computational properties of bird brain nuclei that support this robust behavior. Area HVC is believed to produce a temporal sequence that serves as the metronome for the song. While neural firing in HVC is remarkably stable, we show that small observed variations render the learned song more robust and also allow for more flexible re-learning. In collaboration with the Lois lab, we have explored how song quality is affected under viral perturbations of neuron classes in HVC. While initially the song is dramatically disturbed, it recovers within a few days. We use analysis and modeling to characterize the return of the song and propose mechanisms underlying this recovery.
Invited by Arthur Leblois