John Murray

Strong recurrent interactions in cortical circuits are capable of generating at-tractor dynamics, which may underlie core computational components of cognition, such as working memory and decision making. Working within that conceptual framework, this dissertation explores questions related to dynamics in local cortical microcircuits and in networks of interacting cortical areas.

How do cortical areas differ in terms of their intrinsic circuitry, and how do these differences relate to function? How do the properties of local microcircuits affect the large-scale coordination of multiple areas? In disease states, how does disruption at one level propagate upward to produce dysfunction at another level? To probe questions at multiple levels, this dissertation combines mathematical analysis, computational modeling, and data analysis through collaboration with experimentalists.

Dynamical systems analysis of a simple attractor network model shows how network parameters affect working memory and decision making functions, and how the circuit responds to inputs from other areas. We found tradeoffs inherent in optimizing these functions suggest a role for segregation and specialization in different cortical areas. Modest changes in network parameters, such as the strength of recurrent excitation, may partially underlie the functional specialization of different cortical areas.

We hypothesized that, at the physiological level, differences in microcircuitry across areas may be detectable in terms of differential timescales of neural dynamics. To test this hypothesis, we analyzed of single-neuron spike trains from multiple cortical areas. We found that timescales in the spike-count autocorrelation vary across areas in a manner consistent with anatomical measurements of cortical hierarchy.

Working memory and decision making involve a distributed interacting network of brain areas and cell types, but their differential roles and the nature of their interactions are poorly understood. To examine these issues, we modeled both cognitive functions in a loop network model of interacting modules with specialized properties, and related model dynamics to electrophysiological studies. We also explored decision-making dynamics when a network of two interconnected modules integrates different, potentially conflicting inputs.

The balance between excitation and inhibition is critical for neural function, and disruption of this balance is implicated in neuropsychiatric disease. Using spiking circuit models, we study the effects of cortical disinhibition on neural activity and behavior related to working memory, and compare them to results from pharmacological neuroimaging. Disruption of excitation-inhibition balance within local microcircuits can degrade both local function and long-range coordination between interacting areas.

This thesis contributes to our understanding of the neural circuit basis of functional hierarchies across cortex, the interplay between local and long-range computation in distributed networks, and the utility of computational modeling in the study of neuropsychiatric disease.