Published in Computational and Systems Neuroscience (CoSyNe), Feb 2021.
Auditory perceptual learning reliably enhances cortical representations of task-relevant stimuli in trained animals relative to naive ones. Despite being associated with perceptual improvement, such changes in neural tuning are typically not measured throughout learning. Additionally, there is a large degree of variability in perceptual learning rates across animals, which is largely ignored when recording and interpreting neural activity. To address these limitations, we developed an experimental and computational framework for describing how sensory representations change during auditory perceptual learning. We recorded from a population of neurons throughout the duration of auditory conditioning using two-photon imaging of layer 2/3 excitatory neurons in the auditory cortex of mice. The animals were progressively trained to classify tones as a single, center frequency or non-center through multiple phases of learning. Perceptual discrimination between center and non-center frequencies improved at variable rates, and across individuals, we observed substantial variability in tuning despite similar behavioral performance. Specifically, animals exhibited either a relative enhancement or suppression of the `center' frequency. To make sense of these observations, we trained a network model by reward-modulated Hebbian synaptic learning (Williams, 1992) to solve the same task. We found that the simulated network learns at similar rates as real animals and captures the across-animal variability in tuning. Overall, both data and model reveal nontrivial learning dynamics associated with perceptual learning in auditory cortex, with initial tuning driving across-animal variability in the emerging representations.