Physics Colloquium

In response to sensory stimulation, neurons can generate sequences of complex activation patterns. Yet, neurons in the sensory cortex are active even in the absence of overt sensory stimulation, producing a large amount of 'ongoing,' i.e. spontaneously generated, neural activity that is often indistinguishable from noise. Research in the last two decades suggests that ongoing neural activity may shed light on the architecture and dynamics of neural circuits. Here, I present a new framework encompassing both ongoing and stimulus-evoked neural activity, combining hidden Markov model analysis of neural recordings with biologically realistic models of cortical networks based on spiking neurons. This framework has been applied successfully to the sensory cortex and can be extended to other cortical systems. In the taste system, it has revealed new properties of single neurons and of neural populations, including a reduction of multi-stability and neural dimensionality in response to sensory stimuli, pointing to the existence of local neural clusters (yet to be experimentally confirmed). Using the analytical tools of effective mean field theory, one can explain these properties as emergent features of the network dynamics.