Principal Investigator Michale Fee
We have proposed a new model for premotor temporal control in the song system. The ensemble of RA neurons active at any moment is driven by a small population of RA-projecting HVC neurons that is active at only one time in the RA sequence. As a result, these HVC neurons code for some transient event or ‘state’ in the sequence, rather than longer timescale structure such as song syllables as in the temporal hierarchical model. We suggest that HVC(RA) neurons code for temporal position within the RA sequence. Put in terms of songbird behavior, the motor actions underlying song unfold when the ‘time code’ in HVC acts upon the ‘muscle map’ in RA. The pattern of activation in RA depends on the synaptic connections from HVC(RA) neurons onto RA neurons, suggesting that it is the pattern of HVC(RA)–RA synapses that contains the ‘score’ of the bird’s song. In collaboration with Sebastian Seung's group , we are examining with theoretical techniques the implications of this sparse coding for song learning and for premotor neuronal representations in nucleus RA.