Running Head: CONSTRAINTS ON IMPLICIT STATISTICAL LEARNING Implicit Statistical Learning and Language Acquisition: Experience-Dependent Constraints on Learning

Acquiring spoken language involves implicitly learning the statistical relations among language units. In this chapter, we review recent behavioral and neurophysiological findings from our research group that illuminate the relation of this type of ‘implicit statistical learning’ (ISL) to language acquisition. First, we review evidence illustrating that ISL mechanisms enable the learner to predict upcoming language units. Second, we show modality constraints affecting the operation of ISL mechanisms, with auditory and visual learning biased to learn different types of patterns. Third, we demonstrate that under certain developmental conditions typical ISL abilities are altered, which can detrimentally affect subsequent language acquisition. These findings highlight the experience-dependent nature of ISL and its relation to typical and atypical language acquisition. Introduction Acquiring one’s native language involves at least to some extent implicitly learning the statistical or distributional relations among linguistic units. By “implicit” we simply mean that learning proceeds largely automatically, unintentionally, and often without the learner’s conscious awareness . By “statistical” we mean that language is made up of complex probabilistic associations among language units. There is a growing body of evidence suggesting that this kind of learning is present very early in life, allowing the infant to acquire knowledge about the statistical regularities of language input, which appears necessary for supporting typical language development. In fact, these learning mechanisms are thought to be important not only in language but also in other domains. Especially relevant are situations involving learning items in a sequence: much of an individual’s everyday life requires understanding the order that certain units or