Investigating the conditions under which movie viewers' event models influence their attention

Film, television, and video significantly influence our daily information consumption, making it essential to understand how we comprehend these dynamic scenes and how this affects our visual attention. While many assume that viewers’ understanding influences eye movements during films, research on this relationship is limited. Although a person's event model -their comprehension of events- has been shown to affect their attention during reading comprehension and static scenes, its influence on attentional selection in films and other dynamic stimuli is minimal or non-existent, a phenomenon known as the Tyranny of Film (Davis et al., 2021; Huff et al., 2014; Hutson et al., 2017; Loschky et al., 2015). The current project aims to provide an existence proof of the effect that the event model has on selective visual attention in commercial film clips. Based on the Scene Perception & Event Comprehension Theory (SPECT) (Loschky et al., 2018; Loschky et al.,2020; Loschky et al.,2024), we propose two necessary conditions under which a viewer’s event model could influence their attention to a target: 1) The Differential Event Model Criterion: The target of attention is differentially prioritized by a strong event model when there are clear distinctions in viewers’ event models (specified using event indices); 2) The Default Gaze Probability Criterion: the target of attention has a relatively low default probability of being fixated (i.e., attended to) based on its visual features. We verified whether our selected video clips satisfied the first criterion by assessing their differences in event understanding through linguistic measures and moment-to-moment spatial importance judgment measures. The second criterion was confirmed using a state-of-the-art gaze prediction algorithm, DeepGaze IIE, to ensure that the target’s default gaze probability was relatively low compared to the rest of the scene. Eight suitable video clips that met both criteria were identified through these verification processes. The proposed preregistration experiment will explore whether these differences in event understanding influence attentional selection by tracking participants' eye movements while viewing these clips to predict what will happen next. The eye-tracking study would employ a completely within-subjects design, where participants would view the video clips without context, followed by a second viewing with context. As these eight videos meet our proposed criteria, we anticipate that participants in the context viewing conditions will focus more on the target than in the no-context viewing conditions. We also expect to find evidence for one of our alternative competing sub-hypotheses, suggesting that repeated viewing may either facilitate or interfere with the effect of the event model on attention. This could reveal how viewers reinterpret their prior understanding of an event and how it influences their attention. By accumulating evidence from multiple studies using this systematic approach, we aim for the proposed experiment to provide definitive existence proof for the effect that the viewer's event model has on their attention in commercial film clips. If we observe no effect of event models on attention, it will further support the idea of the tyranny of film in dynamic images, likely due to the constant motion present in films. This work will deepen our understanding of how a viewer’s event model influences their attention during movie viewing.