The Temporal Dynamics of Hierarchical Scene Representations during Scene Categorization

Sammanfattning

Scene perception requires the integration of multiple levels of visual information. The present study examined the temporal dynamics of context-level and schema-level representations during scene categorization and investigated how these representations were related to behavioral performance and task-oriented (indoor vs. outdoor) category decoding. Previously collected EEG data from 39 healthy adults who performed an indoor vs. outdoor scene categorization task were analyzed using multivariate pattern analysis and temporal generalization. Bayesian regression analyses were further used to test whether schema-level and context-level classifier evidence predicted task-oriented category decoding and response time. The results showed that schema-level representations exhibited broader temporal generalization, whereas context-level representations showed a more time-specific pattern. In addition, schema-level evidence was more broadly associated with task-oriented decoding (indoor vs. outdoor), while context-level evidence showed a more temporally constrained association. A similar predictive pattern was observed at the behavioral level: stronger schema-level and context-level evidence was associated with faster response times, with schema-level evidence showing a broader temporal association and context-level evidence showing a more temporally constrained association. Together, these findings suggest that scene categorization is supported by the dynamic coordination between stable schema-level knowledge and more time-specific context-level visual information. Schema-level representations may provide a sustained interpretive framework for organizing scene input, whereas context-level representations may support ongoing updating based on specific visual details.