Relationship between neuronal architecture and variability in cortical circuits

The connectivity of cortical neuronal networks is complex, exhibiting clustered network motifs and ensembles of neurons with high connection probability. However, the significance of these connectivity properties for computation and dynamics in cortex is unclear. In this thesis, I present several studies concerning the behavior of model cortical neurons receiving input from a surrounding network. I begin by studying pairs of neurons, investigating how overlapping excitatory and inhibitory inputs control the statistics of their outputs. I then study fully recurrent networks of neurons with nonuniform connection structures in the form of highly connected neuronal assemblies. These assemblies represent functionally related subsets of neurons, and I investigate their collective behavior in both spontaneously generated activity and evoked conditions. I show that the presence of assembly structure in recurrently coupled, balanced excitatory-inhibitory networks introduces slow timescales in the networks’ dynamics and relate these modeling results to the experimental literature. Next, I present results on how these assemblies form and are maintained with realistic models of synaptic plasticity. In total, these results represent a step toward understanding how connectivity can be modified by sensory experience, and how these changes in turn shape cortical dynamics.