Learning-dependent processing of natural communication sounds in single neurons and neural populations

The ability to learn to recognize new sensory signals such as voices or faces is an important cognitive function in many species. This ability is thought to involve the plasticity of neural representations in high-level sensory cortical areas, but this plasticity is poorly understood. Using European starlings (a species of songbird) trained to recognize natural conspecific song segments, I investigated the emergence of neural representations for learned signals across two auditory forebrain regions : the caudolateral mesopallium (CLM) and the caudomedial mesopallium (CMM). In both CLM and CMM, neurons encoded more information about the motifs (short, stereotyped segments of song) that make up songs paired with reward during training than the motifs that make up novel songs. This shows that behavioral experience is an important modulator of neural encoding in the songbird auditory forebrain. In the natural world, individuals learn which signals convey relevant information for particular behaviors. However, it is unknown how this behavioral information influences neural encoding in the brain. I explored this by training starlings on a paired-motif recognition task where one motif was informative about the behavior required to obtain reward and the other motif was not informative. Following training, single neurons in CLM responded more strongly to informative motifs than to uninformative or novel motifs, whereas single neurons in CMM responded strongly to both informative and uninformative motifs. This suggests that encoding in CLM may serve to emphasize those signals that are particularly behaviorally relevant. Sensory encoding in cortical areas is distributed across many neurons. But how learning alters these neural population representations remains unexplored. To explore this question, I analyzed the correlated activity of simultaneously recorded neurons within CLM. When processing informative motifs, the correlations led to enhanced population discriminability, relative to the correlations when processing uninformative or novel motifs. Thus, the information that a sensory signal conveys about behavior modulates neural encoding in both single neurons and in neural populations. Collectively, these studies demonstrate that behavioral relevance substantially influences neural processing by both single neurons and larger populations in cortical brain region