Abstract
Auditory learning is a key component of vocal learning and communication. Neurons involved in auditory learning are typically examined as single encoders, but there is increasing evidence that the coincident activity of groups of neurons, or 'ensembles', is important for the processing and transmission of sensory information. In songbirds, a forebrain region analogous to mammalian secondary auditory cortex, the caudomedial nidopallium (NCM), is crucial for representing and learning auditory signals. In the awake state, NCM neurons exhibit stimulus-specific adaptation in response to repeated presentation of song stimuli, considered a form of auditory working memory. Yet, how complex stimuli like song are encoded by networks of excitatory and inhibitory neurons is essentially unknown. Using in-vivo single-unit electrophysiology, we systematically unveiled neuronal ensembles that operate at high temporal precision (< 10 ms) across the NCM of awake zebra finches ( Taeniopygia guttata ). Our data show that multiple ensembles are concurrently activated by song with high temporal precision, and that their neuronal composition is heterogeneous, topographically proximate, and biased towards excitatory members. NCM ensembles adapt to song playback and, notably, become more stimulus selective over tens of minutes, accompanied by fast remodeling (membership gain and/or loss) during adaptation. Altogether, our results suggest that song representations in the forebrain can be conveyed by multiple, dynamic network ensembles in parallel. These findings advance our knowledge of the composition, dynamics, and neuronal network reorganization of ensembles as complex sensory stimuli become increasingly familiar.