Assembling Neural Latching Switch Circuits for temporally structured behavior

Abstract

Elaborated temporal structures in behavior have been recognized as a hall-mark of high-level cognitive processes, such as planning or natural language. Systems neuroscience experiments in behaving animals have validated cell assemblies as a fundamental piece of neural circuitry underlying lower-level cognitive processes. Although recent research has identified neural bases for certain temporally structured behaviors, it is still an open question whether a general-purpose circuit - akin to cell assemblies - supports such behaviors. Here we show how Neural Latching Switch Circuits can be assembled as lego-bricks to create neural architectures suited for the implementation of any procedural behavior. By focusing on various forms of temporal structures and leveraging a combination of theoretical tools, we demonstrate how these circuits can be mapped onto interacting brain areas and propose predictions for identifying analogous structures in mammalian brains, for instance in cortical columns. By interpreting neural architectures as computing automata we reveal surprising relationships between behavior and computation.