Sloggett, S. & Dillon, B.
University of Massachusetts Amherst
Reflexives provide an important case study for parsing models which rely on cue-based memory retrieval. Many studies report that reflexives are strongly constrained by syntactic principles (e.g. Principle A) in early stages of antecedent identification (Nicol, 1988; Sturt, 2003; Dillon et al., 2013), a finding which is unexpected from the point of view of cue-based parsing. However, recent work suggests that comprehenders do occasionally consider Principle A incompatible referents which match the reflexive's morphosyntactic features, supporting cue-based parsers (Parker, 2014; Patil et al., 2016). In this work, we suggest that sensitivity to Principle A incompatible referents does not reflect a morphosyntactic feature matching process during antecedent retrieval, but rather a distinct logophoric interpretation of English reflexives. Crosslinguistically, logophors (1) preferentially refer to subjects of speech predicates (Culy, 1994), and (2) are harder to process in the presence of first person pronouns (Huang & Liu, 2001). In two eye-tracking studies, we find that reflexives preferentially access Principle A incompatible referents when (1) such referents are subjects of speech verbs, and (2) a first person pronoun is not present. These findings support the view that logophoricity, rather than morphosyntactic feature matching alone, underlies Principle A fallibility in reflexive comprehension.