Neural networks underlying temporally structured behavior – Cano-T

High-level cognitive processes such as planning, reasoning or language rely on temporal sequences of mental events. Recent studies of animal behavior have characterized the temporal structure of sequences of elementary behavioral events with similar concepts than those used to describe language, such as the notion of grammar. Given the possibility to record neural activity at single-cell resolutions in these animal models, as well as to characterize the anatomy of the corresponding circuits, they offer an opportunity to advance our mechanistic understanding of cognitive processes. Tools from statistical physics of neural networks and dynamical systems theory have proven well suited to account for the emergence of elementary cognitive processes such as working memory or decision making from interactions in structured networks, and have later been successful in accounting for a wide array of experimental data. Here we plan on pushing this approach further, combining it with recent artificial intelligence techniques, to describe how elementary cognitive processes can be combined in a desired temporal order to implement high-level cognitive processes.