The time-course of real-world scene perception: Spatial and semantic processing.

Area of Science:

• Cognitive psychology
• Computational neuroscience
• Visual perception

Background:

• Real-world scene perception is fast but poorly understood.
• Space-centered theory posits rapid recovery of spatial structure from image statistics, enabling semantic inference.
• The relationship between spatial layout and semantic content in scene perception requires further investigation.

Purpose of the Study:

• To investigate the temporal dynamics of spatial and semantic perception in real-world scenes.
• To examine the dependencies between spatial and semantic judgments during scene perception.
• To test the predictions of space-centered theory regarding scene perception.

Main Methods:

• Participants viewed backward-masked real-world scene images for brief durations (13.3–106.7 ms).
• Judgments were made on semantic categories (e.g., beach, road) and spatial structure categories (e.g., open, closed-off).
• Computational analyses were employed to model the perception processes.

Main Results:

• No significant temporal precedence was found for spatial discrimination over semantic discrimination.
• Computational analyses indicated that semantic information is used to discriminate spatial structure categories.
• Findings suggest a reversal of the traditional bottom-up processing model in scene perception.

Conclusions:

• Human scene perception does not rely on a strict bottom-up hierarchy where spatial information precedes semantic information.
• Semantic scene content appears to play a crucial role in discriminating spatial properties.
• These results challenge existing models of visual scene understanding.