Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Investigating ambiguity processing in multi-referent situations: insights from event-related potentials
Poster Session C, Friday, October 25, 4:30 - 6:00 pm, Great Hall 3 and 4
Yi-Chun Ko1, Kara D. Federmeier2, Chia-Lin Lee1; 1National Taiwan University, 2University of Illinois Urbana-Champaign
Establishing coherent referential interpretations is pivotal for language comprehension, given that natural language often presents the challenge of multiple eligible antecedents for an anaphor. In the event-related potential literature, referential ambiguity has consistently been linked with a sustained anterior negativity—the Nref. However, the precise nature of the Nref effect remains elusive. This study aims to adjudicate between two prominent functional accounts, probing whether the Nref effect reflects the detection of ambiguity or the maintenance of potential antecedent representations. We presented stories featuring one, two, or three characters followed by a sentence containing a pronoun. Participants read each story in its entirety and the critical sentence word-by-word, and then answered comprehension questions at the end of each story. This setup created conditions in which the pronoun was either unambiguous (Unambiguous) or temporarily linked to two or three potential referents (2-referent; 3-referent). In the critical sentence, referential ambiguity was resolved with subsequent character names. This design allows us to examine the sensitivity of the Nref effect to referential load during ambiguity and downstream brain responses by time-locking the analysis to the pronoun and the disambiguating name, respectively (e.g., If the rain continues, … (Unambiguous) he / (2-referent) he thinks that Teddy / (3-referent) he thinks that Teddy and Jack … will feel very upset at work tomorrow). At the pronoun, the detection account predicts similar magnitudes of the Nref effect for both ambiguous conditions relative to the unambiguous condition. In contrast, on the maintenance account there should be a greater Nref effect as referential load increases, for the 3-referent versus the 2-referent condition. At the disambiguating name (e.g., Teddy), the detection account predicts no Nref effect because the referent is unequivocal, leaving no additional ambiguity to be detected. In contrast, on the maintenance account there should also be an Nref effect at this position, but only in the 3-referent condition. This is because, although the appearance of the disambiguating name reduces the number of potential referents for the pronoun, there is still more than one potential referent in the 3-referent condition, requiring continued maintenance of multiple referent options. Our results revealed significant Nref effects for both the 2-referent and 3-referent conditions on the pronoun. Surprisingly, the effect was larger in the 2-referent condition. Additionally, an Nref effect was observed in the 3-referent condition upon encountering the first disambiguating name. Collectively, these real-time observations of how referential ambiguity evolves with additional disambiguating information provide intriguing insights into referential processing. The findings are largely, though not entirely, consistent with the maintenance account. Specifically, our results indicate that the Nref may not solely reflect the number of potential participants formally associated with the referentially ambiguous expression. Rather, it may signify participants’ elaborative referential inferences and strategic use of the context to allocate general working memory resources in referential processing and to alleviate working memory demands.
Topic Areas: Meaning: Discourse and Pragmatics, Reading