It is generally believed that behaviors are not mapped to single spikes generated by any one neuron, but rather to groups of spikes. These functional spiking neural activity groups may originate from a single neuron or from populations of neurons firing in synchronic or diachronic manners. The structure of the vast majority of behaviorally relevant neural activity groups is not predetermined by genetics, nor dictated by some sort of an ‘all-knowing teacher’, homunculus. Rather, neural activity groups are formed and modulated throughout life in a dynamic, activity-dependent manner, conforming to evolution and environmental constraints. The formation of neural activity groups is learning; their conservation is memory.
Of the various alternatives, large random cortical networks developing ex vivo are probably the most appropriate experimental model systems for studying the universals governing formation, adaptation, and conservation of neural activity groups. These networks demonstrate extensive functional connectivity and sensitivity of that connectivity to activity. Moreover, the networks are relatively free of predefined constraints and intervening variables. Alternative experimental models (acute in-vivo, or acute in-vitro) allow one to explore ‘what-is-there’, but not ‘how-it-got-to-be-there’. The latter question is tightly related to development.
The body of our network studies implements advanced electronic (multi-electrode array) to interface with large scale developing cortical neural networks. These studies address learning under closed loop settings, adaptation over extended ranges of time scales, stimulus representation, embodiment, dynamics over structure, neuromodulation, impacts of modularity, and more.