Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)




College of Graduate Studies

First Advisor

Jane Joseph

Second Advisor

L. Judson Chandler

Third Advisor

Brian Dean

Fourth Advisor

Andrew Lawson

Fifth Advisor

Thomas Naselaris


There is a considerable and growing interest in the organization and development of neural function at the scale of the entire brain, particularly from activity observed in fluctuations in the blood oxygen level dependent (BOLD) signal in functional magnetic resonance imaging (fMRI). This activity reliably organizes into sub-networks, whether obtained from subjects at rest or from those involved in various perceptual or cognitive tasks. While these networks have often been identified and studied, the precise dynamics involved in their interactions as well as the relationship between organization found in both resting and task-based activity are not well understood. Here I report not only the traditional functional organization of the resting brain as observed through inter-regional correlations, but how this organization changes over time. The dynamics of the brain during rest are not stationary as typically assumed, but vary as different sub-networks co-activate. To produce this more advanced model I apply a new method from the machine learning literature that uses spectral learning to estimate the latent dynamics of brain activity driving these changes. When comparing this model to one from subjects passively viewing faces and other objects, I find distinct changes in a sub-network containing regions of the brain involved in object recognition. Additionally, the importance of fMRI signal variability as an independent within-subject measure has recently gained attention. Looking at those same subjects at rest and passively viewing images, I introduce a simple measure that finds a differential network that may expose the sources of variability driving these dynamics. This network appears to be largely absent in anesthetized primates, and is disrupted in a cohort of cocaine users viewing drug related paraphernalia. These results together suggest a picture of a dynamic brain, with multiple interacting subsystems that are not discrete isolated elements but often overlap. These superpositions of activity may give insights into the dynamics seen during various attentional tasks, where smaller parts of these networks tend to increase in activity. Finally, there may be an independent level of organization that is coordinating this dynamic activity as seen through the variability of these systems over time. This novel network both increases in magnitude and is predictive of age, and discriminates between faces and other objects. These results help further our understanding of network organization in the brain.


All rights reserved. Copyright is held by the author.