To act or be acted upon: Which role is privileged in event perception?

Poster Presentation 53.315: Tuesday, May 19, 2026, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Perceptual Organization: Intuitive physics

Tyler Knowlton1, David Schwitzgebel2, Alon Hafri1; 1University of Delaware, 2École Normale Supérieure – Paris Sciences et Lettres (ENS-PSL)

When entities interact, such as when one hits, kicks, or slides past another, observers rapidly extract which entity acted on which. Such event roles (including Agent or “actor” and Patient or “acted upon”) are central to how events are structured in both vision and language. However, it is unclear whether these roles are prioritized in similar ways across the two systems. Agents are clearly privileged in certain perceptual contexts: They guide attention in visual search and in recognition of actions. In contrast, standard linguistic theories treat Agents as secondary: For example, “the girl kicked the ball” is typically decomposed into (informally) “there was a kicking of the ball, and the girl did it,” a representation in which the Patient (the ball) is more central to the action itself (kicking) than the Agent (the girl) is. Why this apparent asymmetry between vision and language? One possibility is that Agents are actually secondary in visual event representations but Agent advantages nonetheless arise because their posture or movements often reveal what kind of event is being observed (e.g., an outstretched foot unambiguously indicates “kicking”). To test this possibility without postural cues, participants viewed brief animations of two differently colored circles whose spatiotemporal interactions alone revealed role information. They were faster at accurately responding to simple yes/no questions about these relational events (e.g., “Is purple launching orange?”, “Is purple passing orange?”) on trials where the non-Agent (orange) appeared on screen first rather than the Agent (purple). Control experiments ruled out general attentional-shift explanations: The order advantage disappeared for non-relational events (e.g., when two circles simply slid apart in sequence). Together, these results provide support for the idea that vision, like language, assigns primacy to non-Agent roles in the structure of event representations.