The different roles of learning recent and accumulative statistics

We perceive key aspects of familiar environments almost immediately, while perception in unfamiliar environments is slower. In this review, we examine the distinct roles of recent versus accumulative long-term exposure in enabling this efficiency. Accumulative statistics underlie the formation of stable categories (e.g. syllables in our native language), whereas recent events bias our online predictions toward the current context. Typically developing individuals place greater weight on recent events than single earlier events, but also weight accumulative statistics. However, individuals with developmental atypicalities show atypical patterns of statistical learning: individuals with dyslexia tend to assign less weight to long-term statistics, which affects their long-term categories. By contrast, autistics utilize long-term statistics like neurotypicals, but are slower in updating their priors and motor plans by recent events, which reduces their flexibility. These observations suggest that the dynamics of statistical learning impact the strengths and weaknesses of people's social and cognitive skill acquisition.