Testing a mechanistic model for the storage and recall of an auditory memory underlying sensorimotor vocal learning - PROJECT SUMMARY: The ability to imitate others is a hallmark of human cognition and culture, and underlie many aspects of our remarkable ability for thought and language. Songbirds have been established as a powerful model system through decades of research, allowing for a mechanistic understanding of how neural circuit dynamics control the learning and production of complex sequential behaviors. Songbirds acquire their songs through an imitation process reminiscent of human speech acquisition using a well-delineated set of brain regions. By listening to their parents (or tutor sing), young birds form an auditory memory—a template. By comparing their own highly variable babbling vocalizations to this template, young birds gradually refine their own song using a trial-and-error reinforcement learning mechanism that requires a dopaminergic performance signal. Songbird vocal imitations can have remarkable fidelity, often matching the tutor song with striking acoustic and temporal precision. How is this template memory stored, in particular the temporal structure? How is the template compared to the ongoing babbling vocalization? And how does the auditory circuit form an error signal? In our previous work, we found that the timing of song production is controlled by sequential neural dynamics within a premotor region known as HVC. HVC has also been hypothesized to play a role in the formation of the auditory memory used for imitation. This proposal aims to link these two functions within a single framework. We hypothesize that, akin to its role during song production, HVC acts as a clock during tutoring, forming a temporally precise memory of the tutor song that is actively recalled during motor production to guide learning. In preliminary experiments, we have found that thermal cooling of HVC during tutoring causes birds to subsequently develop songs that are faster than their tutor (Aim 1). Furthermore, preliminary experiments show that tutoring produces sequential patterns of neural activity in HVC, which we hypothesize are stored as dynamic sequential ensembles that form the ‘clock’ of the tutor memory (Aim 2). HVC also projects back to the auditory cortex, and we hypothesize that this projection transmits a copy of sparse sequential activity that serves to predictively cancel the auditory signal at each moment in the sequential ensemble. Preliminary experiments have revealed the song-specific adaptation and error responses predicted by this model (Aim 3).
