@

초록

Previous research has shown that the processing of Noun Phrase Ellipsis (NPE) involves the reactivation of structural, semantic, and morphological information associated with the antecedent (Kim et al. 2019), and that reflexive processing adheres to Binding Principle A, requiring a structurally and semantically constrained antecedent (Chomsky 1981, Dillon et al. 2013, Sturt 2003). However, no prior work has examined antecedent retrieval in sentences that contain both ellipsis and reflexive dependencies. The present study tested a strong structural-recovery account, which predicts that sufficiently detailed structural, category-level, and morphological information from the NPE antecedent should guide subsequent reflexive resolution. Using the G-Maze paradigm (Boyce et al. 2020, Forster et al. 2009, Witzel and Forster 2014) and illusions of grammaticality (Wagers et al. 2009) as a diagnostic, we found, contrary to our predictions, that illusions emerged in both grammatical and ungrammatical sentences. This pattern is inconsistent with the predictions of the strong structural-recovery account and suggests that any reactivated structural information does not reliably constrain reflexive resolution in the predicted manner. At the same time, the findings do not isolate a single alternative mechanism and are compatible with multiple explanations, including reliance on linearly local cues, memory-based limitations, or task-related properties of the Maze paradigm. We therefore interpret these results as placing meaningful constraints on strong structural-recovery accounts and as motivating further research to adjudicate among competing explanations.

키워드