Investigação da prosódia e dos padrões de fixação ocular em transições de turno de conversas espontâneas diádicas

Abstract

Conversation is a fundamental social activity characterized by the sequential alternation of speaking turns between interlocutors (Marcuschi, 2003; Sacks, Schegloff e Jefferson, 1974). Studies indicate that prosody plays a crucial role as a cue for predicting turn transitions, providing essential information about the completion of a turn-constructional component and the subsequent transition (Bögels and Torreira, 2015; Gravano and Hirschberg, 2011). This study investigated turn transitions in spontaneous dyadic conversations, focusing on self selection transitions and exploring how prosodic cues — fundamental frequency (F0), intensity, duration, and pauses (Couper-Kuhlen, 1986; Cruttenden, 1986; Crystal, 1969) – and gaze fixation patterns influence the perception and prediction of these transitions. The objective was to understand the interaction between acoustic and visual features to anticipate or react to turn changes. The experiment involved participants listening to excerpts of spontaneous conversations while their eye movements were tracked using eye-tracking methods that followed methodological recommendations from the literature (Holmqvist and Andersson, 2017; Winn et al, 2018). Acoustic analyses were conducted using Praat (Boersma & Weenink, 2024), eye-tracking data were collected with Gorilla Experiment Builder, and statistical analyses were performed in R. Based on the analyzed parameters, the results suggest that prosody contributes to the identification of turn transitions. The analyses indicated the presence of prosodic patterns characteristic of this context. The combination of longer duration and lower intensity at the end of a turn with higher intensity and shorter duration at the beginning of the subsequent turn creates a perceptible contrast that helps interlocutors recognize turn transitions. Variations in intensity and duration modulation act as acoustic and temporal cues, contributing to reduced overlaps or gaps and enhancing predictability and interaction fluidity. Eye-tracking results indicated that participants tended to exhibit anticipatory gaze patterns toward the next speaker during the pre-transition window (PreTW). During the transition window (WOI), fixations were distributed between the current and next speakers, while in the post-transition window (PostTW), a greater proportion of fixations were directed at the previous speaker, suggesting a reactive visual pattern influenced by specific prosodic cues. This study expands the understanding of turn transitions in Brazilian Portuguese, a language still underexplored in this field (Barbosa, 2012). Moreover, the findings offer practical implications for the development of automated conversational systems, such as virtual assistants, by proposing strategies based on prosody and visual attention to improve human-machine interaction (Skantze, 2021). Additionally, the results can be applied in areas such as education, speech therapy, and intercultural communication, providing a foundation for future research on the cognitive processes involved in speech perception and turn transitions (Barry e Andreeva, 2001; Ryngaert, 2018; Santos e Silva, 2021).