A systematic investigation of prediction error and learning in infancy

Abstract

Prediction, or the ability to use past experiences to generate expectations about future sensory input, is increasingly believed to be crucial for adult cognition and learning. Despite decades of targeted research on prediction in adults, it remains unclear about how prediction operates during infancy. In Chapter I, I discuss how pupillometry, or the measurement of pupil size, can be used to investigate prediction and its component parts, namely anticipation, expectation, prediction error and updating, with an emphasis on prediction error. In Chapter II, I take a behavioral (measuring infants’ pupil size) and computational approach (using an errordriven learning model, the Rescorla-Wagner Model) to understand how infants use prediction error and whether this ability changes across the lifespan. Chapter III describes a series of studies that uses pupil size to measure prediction in a standard visual statistical learning task. Chapter IV sought to combine eye-tracking and pupillometry in order to understand how prediction error impacts infants’ ability to predict, or elicit anticipatory eye movements, in the context of the Visual Expectation Paradigm. Lastly, Chapter V is a follow-up investigation that also uses eyetracking and pupillometry, to determine whether prediction error is the main driver behind infants’ ability to update. Taken together, these studies expand on the theories that have proposed that prediction is a core aspect of adult brain function to human development and specifically investigates how prediction and prediction error supports learning in infancy.