Natural Acoustic Signals and their Neural Representation

Acoustic communication signals underlie a wide range of perceptual and cognitive behaviors, and typically drive strong, selective neuronal responses in higher cortical regions.  As such, they provide attractive targets for studying the neural mechanisms of real-world auditory processing and cognition, and information transmission in natural systems more generally. But natural signals are difficult to work with. Their spectral and temporal complexity can be difficult to quantify, parameterize, and model; and their high-dimensional structure challenges many classical notions of stimulus encoding. I will discuss recent studies from my lab that address these challenges, describing a suite of unsupervised machine learning techniques that permit direct measurement, parameterization, and generative control over complex natural signals, and introduce a topological technique to analyze activity in arbitrarily large neural populations with single trial precision.  I then discuss recent experiments in European starlings, a songbird species, that apply these techniques to understand neural mechanisms supporting perception and cognition of natural communication signals.