Changes in statistical learning across development

Abstract

Statistical learning enables learners to extract the environmental regularities necessary to piece together the structure of their worlds. The capacity for statistical learning and its properties are likely to change across development from infancy to adulthood. Acknowledging this developmental change has broad implications for understanding the cognitive architecture of statistical learning and why children excel in certain learning situations relative to adults. In this Review, we first synthesize empirical work on the development of statistical learning, which indicates that it improves with development only for certain forms of input. Taking inspiration from related cognitive and neural findings, we then consider developmental changes in the properties of statistical learning. Infants and young children might have a broader and less-directed curriculum for learning and represent the outcomes of learning differently from older children and adults. This synthesis offers insight into how developmental changes in statistical learning from infancy through adulthood might fundamentally alter how children interact with, learn about, and remember their experiences. From infancy, humans learn the regularities of their world using statistical learning. In this Review, Forest et al. consider how statistical learning changes quantitatively and qualitatively across development, considering influences on the input to learning and the resulting memory representations.