Infant fMRI: A Model System for Cognitive Neuroscience

Date of Award

Spring 2021

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Turk-Browne, Nicholas


Studying the infant mind and brain promises to benefit three distinct fields of science in critical ways. In developmental science, understanding infancy is necessary for appreciating the foundations of the human life and determining what changes with development. In cognitive science, revelations about infant cognitive capacities can constrain theories by limiting the possible explanations of how those capacities emerge. In neuroscience, infants offer the unique possibility to study large-scale neural change in humans, while also revealing principles that underlie the optimization of neural circuitry. Yet, there are limits on the kinds of questions that can be asked by the methods available to study the infant mind. Traditional behavioral measures, like looking time, are low-dimensional and require tight experimental control to reach conclusions. Recently, surfaced-based neuroimaging methods, like electroencephalography and functional near-infrared spectroscopy, have gained prominence. These are powerful tools, providing excellent temporal precision or enabling naturalistic testing, respectively. By contrast, functional magnetic resonance imaging (fMRI) is able to ask questions about the infant brain that surface-based methods cannot. fMRI can evaluate deep-brain structures, like the hippocampus, and can localize neural activity precisely. Moreover, like other neuroimaging methods, fMRI provides rich, high-dimensional data which itself can be used to indirectly measure cognition. Resting state fMRI with asleep infants has become more common recently; however, task-based fMRI remains the only way to ask questions about many aspects of cognition. Despite these benefits, task-based fMRI remains a challenge due to infant movement, fussiness, and sleepiness. This might explain the small number of fMRI studies that have been published in the last 30 years that report data from awake infants. In this thesis I describe four experiments that probe the foundations of the infant mind with task-based fMRI. First, I describe the technical and practical innovations I have developed in order to collect high quantities of fMRI data from awake infants. I further show evidence that these methods are able to acquire high quality data at the individual participant level across two cohorts. These methods establish a viable approach to conduct adult-grade cognitive neuroscience with awake behaving infants. Second, I studied the organization of the infant visual system. The presence of retinotopic organization was tested across 17 sessions in infants between 5 and 23 months. I found evidence of retinotopic organization even in my youngest participant, suggesting that retinotopic organization does not emerge during infancy. I observed subtle changes in region size with age, but overall the evidence suggests surprising maturity of the visual system in terms of its mesoscale organization. Since visual perception is developing over infancy and childhood, changes in retinotopic organization are unlikely sufficient to explain changes in vision. Third, I investigated how the infant brain supports the orienting of attention to salient stimuli. I used a cuing paradigm in which infants were either correctly or incorrectly cued to the target location and evaluated both the time it took them to saccade to the target and the evoked neural response to each trial. I found robust behavioral evidence of attention in all 24 sessions (infants aged 3--12 months old). I also found evidence that frontal regions like the anterior cingulate cortex and medial frontal gyrus were recruited to support attention orienting. This indicates that frontal regions are actively supporting cognition, even early in infancy. Fourth, I tested how the infant brain supports learning. I used a statistical learning paradigm in which infants were shown stimulus sequences that either contained structure that they could learn over time or sequences that did not contain structure. Our specific focus was on whether the hippocampus would show evidence of learning. Across 24 sessions in infants aged 3 to 24 months I indeed found evidence that the hippocampus was more activated by the sequences containing structure, but only after an initial period of exposure. In fact, the timeline of this learning was consistent with behavioral evidence from infants and the localization of this was consistent with adults. Hence, the hippocampus is recruited to support learning in early infancy. Together these projects establish a new approach to developmental cognitive neuroscience and provide critical insight into the infant mind.

This document is currently not available here.