Disfluências como pista de fluência na leitura em voz alta

Abstract

This thesis investigated disfluencies in reading aloud as cognitive and behavioral indicators of reading processing, understanding them not only as performance failures but also as cues to fluency and the automatization of decoding. Situated within the field of the Cognitive Psychology of Reading, the research is grounded in informationprocessing models, with emphasis on the Dual Route Model (Coltheart et al., 2001), articulating pedagogical and clinical perspectives on reading assessment. The study is part of the project “Impacts of the COVID-19 Pandemic on the Language of Children and Adults” and is based on a secondary analysis of a longitudinal and multimodal database, composed of four data collections over one academic year, involving students from the 6th to the 9th grades of elementary school at the Colégio de Aplicação of the Federal University of Sergipe (CODAP/UFS). Oral reading sessions were audio-recorded, transcribed, and analyzed in terms of reading rate (WCPM), overall disfluency rate, and specific types of disfluencies (pauses, hesitations, repetitions, self-corrections, truncations, omissions, and substitutions). The results indicated a significant effect of data collection on WCPM, with a progressive increase in fluency over time and across grades, and an overall reduction in the disfluency rate, although with intra- and inter-grade variability. Truncations and guessing were more frequent in the earlier grades and among less fluent readers, showing a tendency to decrease with advancing schooling and reading automatization. In the second study, an exploratory investigation of perceptual judgment of reading aloud was conducted, with the development and validation of an instrument whose overall Content Validity Index was 0.802, indicating adequate content validity. Inter-rater agreement for the overall fluency score was moderate at the individual level (ICC(2,1) = 0.483) and high when the mean of the judges was considered (ICC(2,k) = 0.924), demonstrating collective consistency despite variability among raters. There was no significant effect of professional area on overall fluency scores (p = 0.621), nor on scores for pauses, hesitations, repetitions, self-corrections, omissions, and substitutions (p > 0.05 in all cases), indicating that speech-language pathologists, teachers, and psychologists used similar criteria in these measures. In contrast, agreement for the classification of reading profiles was low (κ = 0.228), suggesting greater interpretative instability in categorical decisions. Taken together, the results support the interpretation of disfluencies as functional markers of fluency development, associated with the dynamic balance between phonological and lexical routes, and indicate that reading judgment is more stable in continuous measures than in profile classifications. These findings contribute to a non-pathologizing approach to reading aloud and provide support for more precise assessment practices in educational and clinical contexts.