Does prediction error explain referential adaptation?

Pronoun resolution is guided by various discourse cues. For instance, ambiguous pronouns are often assigned to the subject of the preceding sentence (e.g., Ana went shopping with Liz. She bought a hat ). Studies have found that these discourse biases can be learned through exposure by tracking frequent pronoun-reference patterns (e.g., pronouns often refer to subjects), a process known as referential adaptation (e.g., Johnson and Arnold, 2023 ; Kaiser, 2009 ; Ye and Arnold, 2023b ). Yet, the mechanisms underlying this learning remain unclear. Two experiments examined whether referential adaptation is driven by prediction error, which has been suggested to explain syntactic adaptation (e.g., Chang et al., 2012 ). This account predicts that learning depends on how surprising the input is, with more unexpected input leading to stronger adaptation. We tested this prediction by exposing people to discourse patterns within both contexts where the reference was highly expected (Exp. 1) and equi-biased (Exp. 2). Robust referential adaptation occurred in both experiments. These findings show that strong expectation violations are not necessary for referential adaptation and suggest that adaptation is not straightforwardly determined by the strength or consistency of prediction error.