Harvard University -- Department of Organismic and Evolutionary Biology -- Center for Brain Science
The Neurobiology of Vocal Learning
The research in our lab focuses on the principles and mechanisms used by neural circuits to generate complex, learned behaviors. We are using the songbird as a model system, concentrating our efforts on understanding the process of vocal learning. The zebra finch, our species of choice, acquires its song in much the same way that we learn many of our motor skills – by imitating a tutor. It first listens and memorizes the tutor song, then proceeds with trial-and-error vocal experimentation, producing highly variable song. By continuously evaluating its own performance relative to the memorized song template, the songbird slowly converges on a copy of the tutor song. The ultimate goal of our research is to arrive at a mechanistic description of how this learning algorithm is implemented by neurons and their connections. To this end we manipulate, measure from, and model the neural circuits involved in song learning.
The process of song learning and song production is thought to be governed by a set of interconnected brain nuclei, known collectively as the song circuit. Our research is aimed at understanding how this circuit implements its function.
Manipulating the song circuit
By observing the effects of targeted manipulations on singing, we can infer the role and function of given brain regions and neural mechanisms. Ways in which we perturb the song circuit include pharmacology, electrical stimulation, and lesions. A recent example of this approach is the inactivation of LMAN, the output nucleus of a basal ganglia circuit that projects to motor area RA. By infusing a blocker of neural activity into LMAN in the juvenile learning bird, the otherwise variable song became stereotyped. Thus, a simple perturbation made it possible to infer the function (induction of song variability) of a given brain region (LMAN) relative to the process of song learning. We are now in the process of expanding our toolbox to include molecular techniques for manipulating brain circuits.
Measuring from the song circuit.
We use miniature, motorized electrode drives to record activity from single cells in targeted brain regions of the singing bird. These recordings allow us to understand how the song is represented in different parts of the song circuit and how this neural representation evolves as a function of learning. We also combine electrical recordings with manipulations of nearby circuits to shed light on how neural responses in a given area are influenced and shaped by activity in adjacent brain regions.
Modeling the song circuit
Our experimental observations inform a theoretical model of song learning and song production. Our model incorporates known anatomy, circuit mechanisms and biophysics, and serves to distill our observations into their essence as well as to generate testable hypotheses. We hope that the iterative process of creating hypotheses, testing them, and using the results to update our model will ultimately lead us to a mechanistic and biophysically plausible description of how a neural circuit (the song system) can learn to produce a desired motor output (song).