Anticipatory processes can influence how quickly comprehenders can process novel linguistic input and how they learn from linguistic surprises. This chapter outlines experimental evidence establishing the psychological reality of anticipatory processes and sketches some contemporary accounts that explain how comprehenders generate predictions from linguistic input. Accounts like Pickering & Gambi’s (2018) formulation suggest that comprehenders covertly engage language production mechanisms to generate predictions about future input and to know when it is time to stop processing current input. Kuperberg and colleagues’ (2021, 2023) formulation lays out a multi-layered network that produces predictions for several different types of linguistic and semantic information (phonological/orthographic, syntactic, lexical, event). N-gram accounts (Brennan, 2020; Hale, 2003, 2016) focus on word predictions and include formal metrics of entropy and surprisal derived from information-theoretic frameworks like Shallice’s. On this account, comprehenders store in long-term memory strings of words (N-grams) and these stored patterns serve as the basis for calculating entropy (how many different continuations are possible at a given point) and surprisal (how likely is a specific word in a specific context). We present a variety of evidence indicating that n-grams may not be the sole or main basis for predictions.