Prasanth Chandran¹, Baile Dreyfuss², Camille Dortland², Claire Holt², Eva Naab², Karla Salto², Laurel Conner², Maddy Pilarz², Joseph P. Magliano³, Lester C. Loschky⁴; ¹Doctoral Student, Kansas State University, ²Undergraduate Student, Kansas State University, ³Professor, Georgia State University, ⁴Professor, Kansas State University

A person's event model—their understanding of events—has been shown to influence their attention during reading and viewing static scenes. But its effect on attentional selection in films is minimal or absent, a phenomenon known as the Tyranny of Film. This study provides evidence for the event model's effect on selective visual attention in conventional film clips. Using the Scene Perception & Event Comprehension Theory (SPECT), we hypothesized two necessary conditions for a viewer’s event model to impact their attention to a target: 1) The Differential Event Model Criterion: The target of attention is prioritized by a strong event model; 2) The Default Gaze Probability Criterion: The target of attention has a relatively low visually-based default gaze probability. We verified whether our selected video clips met the first criterion by examining differences in event understanding through linguistic measures and moment-to-moment spatial importance judgments. We confirmed the second criterion, that the target’s default gaze probability was relatively low, using a state-of-the-art gaze prediction algorithm, DeepGaze IIE. We identified eight video clips fulfilling both criteria. We then tested whether these differences in event understanding affected attentional selection in a within-subjects eye-tracking study (N = 44) where participants first watched the videos without context, then with context. The target of attention was fixated four times more often in the Context condition than the No-context condition, which was at chance level. Additionally, over time in the Context, the likelihood of fixating on the target increased as its importance grew; this increase was five times larger than the No-context condition. Finally, a priori event-model importance had a stronger influence on attentional selection than a priori default gaze predictions based on stimulus features. In conclusion, these two criteria are sufficient for the event model to influence attentional selection during film viewing, thereby overcoming the Tyranny of Film.

Acknowledgements: Research reported in this abstract was partially supported by the Cognitive and Neurobiological Approaches to Plasticity (CNAP) Center of Biomedical Research Excellence (COBRE) of the National Institutes of Health under grant number P20GM113109.