In order to understand the origins and development of our remarkable cognitive abilities, developmental scientists develop clever experiments to understand what infants see, think, and understand. In these experiments, infants eyes are tracked while they watch a series of carefully designed videos. The patterns of eye movements they generate can then be used to draw inferences about the unobservable cognitive process in the infants' heads which generate this eye gaze behavior. Unfortunately, these inferences are made quite difficult because of the complexity of the infants' occulo-motor systems. Eye movements are affected by many variables other than those under the experimenter's control. This can make searching for the meaning in this data much like searching for a needle in a haystack. This summer, Dr. Shohei Hidaka and I developed a novel statistical analysis framework for reducing the size of this proverbial haystack. We subsequently tested this analysis in a series of computational simulations, showing it to be capable of dealing with three well-known problems in the current literature. It can automatically discover the number of qualitatively different groups of infants in the data, discover the contribution of multiple interacting processes to the observed eye movements, and discover non-linear learning functions if they best account of the generated gaze data. We then applied this analytic framework to data collected by Dr. Rachel Wu, a colleague at Birkbeck, University of London. Analyses of these data uncovered processes which were invisible to standard statistical analyses, and which shed light on how infants learn to use social cues to learn about their world.