Modelo probabilístico na interpretação de pronomes possessivos em contexto de transferência de posse

Abstract

Language processing seeks to understand how the speakers of a given language comprehend it, that is, the process by which speakers interpret their language. In the field of psycholinguistics, processing is widely used to study anaphora, which is a way of referring to a previously mentioned term. The pronoun, in turn, is one of the most common ways to accomplish anaphora, given that it tends to retrieve the most salient entity of the discourse. In a sentence like “José transferiu/estava transferindo para Luciana o dinheiro dela/dele bem rápido hoje cedo.”, it is observed that the possessive pronoun (dele/dela) refers to one of the previously mentioned entities (the subject “José” or the object “Luciana”). This study employed two experiments to explore the interpretation of possessive pronouns by speakers. This study employed two experiments to explore how speakers interpret possessive pronouns. Experiment I used self-paced reading to examine, through reading times, whether participants found it more difficult to retrieve the subject or the object of the sentence depending on the verbal aspect (perfective or imperfective), and whether men and women use the same strategies to interpret possessive pronouns. Experiment II used the same sentences as Experiment I, but within a sentence-completion paradigm. According to previous experiments (Rohde, 2008; Godoy e Mafra, 2018), for verbs in the perfective aspect, processing anaphoras with objects as antecedents will be less costly than with subjects as antecedents, while the opposite occurs with imperfective verbs. The initial purpose of this study is to analyze the interpretation of possessive pronouns, comparing it to other pronouns and referential expressions, including in languages other than Brazilian Portuguese. The second goal is to verify whether the Kehler and Rohde Theorem (Rohde and Kehler, 2013), a probabilistic model, can predict anaphora resolution. However, our results showed no influence of verbal aspect on retrieving certain antecedents in either Experiment I or Experiment II. Nonetheless, in Experiment I, we found that sentences in the perfective aspect tended to be processed more easily than those in the imperfective aspect, regardless of the antecedent. Additionally, when analyzing the data for female participants separately, their responses partially supported our hypothesis.