Idiosyncrasies of processing plural pronouns

Author(s)

Mahayana Cristina Godoy

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Instituto de Estudos da Linguagem
• Defense date: 2014-03-13

Examining board members

Edson Françozo; Ruth Elisabeth Vasconcellos Lopes; Aline Alves Fonseca; Elisângela Nogueira Teixeira; Renato Miguel Basso

• Advisor: Edson Françozo

Abstract

The study of pronoun resolution has traditionally relied on the presupposition that the parser starts searching for an antecedent as soon as it is confronted with a pronoun. This view has guided most of the large amount of work on the processing of singular pronouns (e.g. Arnold et al., 2000; Sturt, 2003) as well as the scarce work that has been done on the resolution of plural pronouns (e.g., Oakhill et al., 1992). Here, we study some occurrences of plural pronouns that seem to defy such an incremental view, as in "Alice used to eat beef everyday, and they usually prepared a delicious New York strip at the steakhouse". In this text, a referent for the plural pronoun "they" is inferred through the locative "at the steakhouse", but it is not until the end of the sentence that the reader has this information. These observations raise the questions of (i) whether plural pronouns create an immediate co-referential relation with a possible antecedent and (ii) whether processing plural pronouns really requires an antecedent by the time these expressions are read. In an eye-tracker experiment, we tested how quickly singular and plural pronouns build a co-referential link with a referent that was explicit in a previous sentence. Our results show that singular pronouns are immediately read as co-referential to their antecedent, and canceling this co-reference evokes a greater cost in comparison to situations in which the co-reference is kept through the whole sentence. On the other hand, plural pronouns does not seem to create such an immediate relation with its supposed antecedent, and canceling the co-reference does not result in extra processing cost. In a self-paced reading task, we tested how crucial it was for the processing of plural and singular pronouns that these expressions had an antecedent. Singular pronouns with no antecedent showed greater reading times in comparison to pronouns that had an antecedent. Reading times for plural pronouns were the same regardless of the presence of an antecedent. From the perspective of pronoun processing, we can conclude that the results described above do not fit the presupposition that singular and plural pronoun processing follow the same resolution strategy. Because plural pronouns may depend on information presented later in the discourse, its resolution may be delayed without causing extra processing cost. Furthermore, because our data suggest that plural pronouns may continue unresolved, these results may also be interpreted as evidence that language processing, in specific situations, may rely on superficial representations of discourse structure (AU)