Statistical regularities are abundant in language. Short-range patterns at different levels of language have been shown to facilitate a reader’s text processing (i.e., orthographic, phonological, syntactic). However, the impact of long-range statistical dependencies between words, making up the context, remains understudied, leaving several important questions unanswered. How is people’s reading behaviour influenced by context, allowing for predictions of the next word? What role does contextual predictability play in the development of reading fluency? How do adult readers balance the trade-off between global and local regularities while reading? By utilising natural language processing models, we will be able to overcome past difficulties in measuring predictability and quantify long-range word predictability in a ‘continuous’ way. In three complementary work packages we will investigate how context-facilitated predictability affects both reading fluency and reading comprehension during natural reading at different stages of reading development. In addition, this project will examine how experienced readers are able to flexibly adapt to ‘local’ long-range regularities in text with diverging patterns. Combined, these efforts will deliver new insights into how reading behaviour during development and in different situations is influenced by context.