Attentional synchrony in films: A window to visuospatial characterization of events

Suchan, Jakob · 2022 · OpenAlex

DOI: 10.1145/3548814.3551466

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates how visuospatial attributes in film scenes influence human attention and event comprehension. The authors address the research question of how specific visual cues—such as motion, gaze, and body pose—drive attentional synchrony among viewers. Motivated by the need to understand event segmentation and prediction in everyday activities, the paper proposes that correlating attentional measures with visuospatial features provides a window into how observers characterize events. The work aims to formally represent everyday interactions through a visuospatial model and demonstrate how these descriptions explain observer attention mechanisms in narrative films. The researchers developed a comprehensive visuospatial model that categorizes scene elements into static and dynamic types and defines scene structure through attributes like visibility, presence, motion, spatial position, human action, head movement, gaze, hand action, and body pose. This model serves as a semantic ground truth for annotating multimodal human interactions. To evaluate this model, the team selected ten film scenes from a dataset focused on qualitative spatio-temporal analysis. They collected eye-tracking data from 32 participants per scene using a Tobii X2-60 Eye Tracker at 60 Hz. Expert human evaluators annotated the scenes and corresponding eye-tracking data using the ELAN tool, mapping visual events to the controlled vocabularies of the visuospatial model. This process ensured high-quality, uniform annotations linking low-level fixation data to high-level object-level attention. The analysis focused on attentional synchrony, defined as the percentage of viewers looking at the same region or body part simultaneously. The results demonstrated that attentional synchrony varies significantly based on visuospatial complexity. Specifically, the study found higher synchrony when characters were alone in the frame or engaged in specific, distinct behaviors, highlighting reactive and anticipatory gaze scenarios. By segmenting scenes into high (>50%) and low (<50%) synchrony periods, the authors characterized the distribution of attention in relation to visuospatial features. This allowed them to tease apart event segments and provide an indirect measure of attentional saliency, showing how specific cues direct observer selection within a given context. The significance of this work lies in its contribution to cognitive science and AI systems. By providing a context-agnostic semantic interpretation of multimodal human behavior, the study offers a framework for automated processing of both low-level features (e.g., motion) and high-level features (e.g., referential gaze). The findings imply that visuospatial characterization can serve as a robust method for knowledge representation, visual sense-making, and declarative reasoning in artificial intelligence. This approach bridges the gap between low-level visual data and high-level semantic understanding, offering a structured way to model how humans perceive and interpret dynamic, multimodal environments.

Key finding

Formal visuospatial characterization of film scenes correlates with observer attentional synchrony, allowing for the prediction of attentional selection based on specific cues like motion and gaze.

Methodology

lab_experiment

Sample size: 320

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-29.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• attention allocation