Newport, E.
In recent years a number of problems in the brain and cognitive sciences have been addressed through statistical approaches, hypothesizing that humans and animals learn or adapt to their perceptual environments by tuning themselves to the statistics of incoming stimulation. However, statistical learning is not merely learning the patterns that are presented in the input. Our research also shows that there are maturational changes in statistical learning, with children sharpening the statistics and producing a more systematic language than the one to which they are exposed. Our most recent work examines variation in relation to linguistic universals, suggesting that, when inconsistencies occur on dimensions on which languages tend strongly to align in one direction, learners also shift the languages they learn in this direction, and children do this even more strongly than adults. These processes potentially explain why children acquire language (and other patterns) more effectively than adults, and also how systematic language structures emerge in communities where usages are varied and inconsistent. The central lessons for statistical learning approaches are that optimal statistical learning is not always veridical learning of the input statistics; and that children do not necessarily learn identically to adults.