Abstract

Cortical computations are carried out by the concerted dynamics of neural circuits comprised of a large number of neurons. While we have extensive knowledge of the cellular-level properties of cortical neurons, we know very little about how they contribute to circuit-level computations. We have developed a theoretical framework based on the hypothesis that a wide range of these properties serve the purpose of allowing robust dynamics in a circuit in the face of the severe informational bottleneck imposed by spike-based communication between the constituent neurons of the circuit. This theory accounts for different forms of short-term plasticity as well as the well-known nonlinear integrative properties of dendrites. In particular, our theory allowed us to predict properties of short-term plasticity in a synapse and dendritic integration in the postsynaptic neuron based on the in vivo statistical patterns of activity in the presynaptic cell or population. Experimental data, available in the literature and provided by our collaborators, has confirmed these predictions. This work thus offers a principled link between different levels of neural organisation most often studied in separation under the rubrics of “cellular” and “systems” neuroscience.