Nayoung Kim¹, Oakyoon Cha¹; ¹Sungshin Women's University

Relational representations can be useful in understanding a visual scene. For example, recognizing an object that provides physical support (e.g., table) to another object (e.g., book on the table) may facilitate scene understanding. Hafri and colleagues (2024) demonstrated that participants recognized a scene faster when the reference object (table in the example) was presented before the figure object (book in the example) than the reverse order of presentation. Even for species lacking a neocortex, relational information can guide their behavior in response to environmental challenges. In the present experiment, we investigated the possibility that subcortical areas contribute to the building of relational representations. To address this possibility, we used a dichoptic presentation technique. When two objects are presented to different eyes, the two objects are processed separately in the subcortical visual processing areas before information from the two eyes integrates at the primary visual cortex. If subcortical areas contribute to relational representations, dichoptic presentation should disrupt these processes. Participants first read a sentence and then viewed scenes depicting a reference and a figure object, judging whether the sentence accurately described the scene. Reference and figure objects could be presented to both eyes (dioptic condition), to the same, one eye (monoptic condition), or to different eyes (dichoptic condition). Object presentation order (reference first or figure first) was also manipulated. Across all conditions, participants responded faster when the reference object preceded the figure object , replicating previous findings. The effects of presentation order of objects were comparable across dioptic, monoptic, and dichoptic conditions (BF < 1/3). These results suggest that relational representations rely predominantly on cortical processing rather than subcortical pathways, at least when it is judged based on the relational representations derived from language-based descriptions.

Acknowledgements: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science and ICT) (No. RS-2023-00211668).