Research over the last two decades has established that human infants, children and adults are highly sensitive to the statistical structure of their environments, including many layers of language structure. However, the mechanics of how statistical learning operates and develops are still unclear. I will discuss three studies that each aim to get “under the hood” of early statistical language learning to better understand how infants collect and process distributional information. Each of these studies connects infants’ learning outcomes with their behavior during the learning process. Two studies examine how infants’ attention skills are related to their ability to use distributional information to segment words from audiovisual speech (Study 1) and infer word meanings (Study 2). Study 3 employs a spatial prediction task to investigate the process by which infants build representations of predictable structure. Together, the findings from these studies demonstrate that characteristics of the individual learner, such as their attention skills and prior experience, shape how infants extract statistical language information.