I am fascinated by the myriad of statistical regularities in spoken and written language, whether these regularities assist the incredible feat of acquiring language, and the nature of the learning mechanism(s) which might underpin sensitivity to these regularities. In this talk I will share my discovery of a rich source of probabilistic orthographic cues to lexical stress in English and in other European languages. Of course, the discovery of such regularities does not mean that children actually use them when learning to read. Another part of my research program is devoted to investigating whether these kinds of cues are used and how they are learned. I will discuss my research on how children learn to assign lexical stress during reading aloud via a triangulation of methods: corpus analysis of probabilistic cues to lexical stress in words that are found in children’s age-appropriate storybooks, behavioral testing of children aged 5-12 years of age to test sensitivity to these cues, and connectionist computational modeling to simulate children’s reading aloud. Broadly speaking, it is thought that a form of implicit learning, such as statistical learning, underpins sensitivity to these and other kinds of statistical regularities when children are acquiring spoken and written language. I will discuss some of my research on individual differences in children’s capacity for statistical learning and on the link between these individual differences and reading ability.