O processamento anafórico de surdos na leitura do português brasileiro

Abstract

This dissertation investigates the processing of anaphoric direct objects during the reading of Brazilian Portuguese (BP) by deaf individuals in a bimodal bilingual context, users of Brazilian Sign Language (Libras) and written BP, taking hearing readers as a comparative reference. In light of experimental psycholinguistics, the study focuses on the mechanisms involved in maintaining and resuming referents, which are fundamental to textual coherence, such as working memory, information load, and discursive integration (Almor, 1999; Gordon; Grosz; Gilliom, 1993; Leitão, 2022). Two strategies of coreferential resumption in the direct object position are analyzed: lexical pronouns (PR) and repeated nouns (NR), based on the Information Load Hypothesis (ILH) and the Repeated Name Penalty (RNP), according to which more informationally dense expressions tend to impose a higher processing cost (Almor, 1999; Gordon; Grosz; Gilliom, 1993). Considering the specificities of bimodal bilingualism and the modal differences between Libras, which is visually-spatial, and written BP, which is linear, we start from the hypothesis of a higher cost associated with NR resumptions in L2 reading (Gadelha, 2012; Sendek et al., 2023; Silva, 2025). Methodologically, a comparative experimental design was adopted between deaf and hearing readers, using the non-cumulative self-monitored reading technique to measure online processing (Leitão, 2005; Souza; Franzen; Schlichting, 2019). Adults with a minimum of higher education participated, characterized by self-declared proficiency and the Brazilian Portuguese Lexical Verification Test (TVLPB) (Sá et al., 2021), whose results showed wide variability in proficiency among the deaf. The independent variables included type of resumption, participant profile, and mode of participation, while the dependent variables included reading times, response times to comprehension questions, comprehension question accuracy rates, and proficiency rates. Multiple linear model analyses indicated combined effects of linguistic and individual factors on reading times. The Wilcoxon test revealed differences between resumption conditions in specific groups, while the chi-square test did not indicate significant associations in most comparisons, suggesting greater instability in performance only in the GE (Libras-L1/PB-L2) - face-to-face. In general, listeners had lower reading times and greater stability, with clear evidence of PNR only in the GC (PB-L1/Libras-Ø) - face-to-face. Deaf participants had higher reading times and greater variability, with a tendency toward PNR, especially among participants with greater proficiency. The results contribute to the understanding of anaphoric processing in BP by deaf individuals, from the perspective of PNR under the effect of multifactoriality, broadening the scope of experimental literature by incorporating a population that has been little investigated, offering theoretical and applied subsidies for bilingual education of deaf individuals (Quadros, 2004, 2005, 2006; Karnopp, 2004; Ferreira Brito, 2010).