A large positive hysteresis effect for scene categories.

Area of Science:

• Cognitive psychology
• Neuroscience
• Computer vision

Background:

• Seamless visual perception relies on balancing recent history and current sensory input.
• Perceptual hysteresis describes the visual system's tendency to favor past (positive hysteresis) or present (negative hysteresis) information.
• The dominant direction of hysteresis in complex, changing naturalistic scenes remains largely unexplored.

Purpose of the Study:

• To investigate the direction of perceptual hysteresis during the recognition of semantic categories in dynamically changing naturalistic scenes.
• To determine whether the visual system exhibits positive or negative hysteresis when interpreting continuous visual transitions between scene categories.

Main Methods:

• Utilized generative adversarial networks to create smoothly transitioning scene-image sequences between distinct categories (e.g., bedroom, living room).
• Presented sequences in opposite directions to participants, who identified the perceived category change point.
• Analyzed response data against the true category boundary, controlling for motor response time and prior expectations.

Main Results:

• Perceived category boundaries were systematically delayed relative to the true boundary, indicating a bias towards the recently perceived category.
• A robust positive hysteresis effect was observed across all transition directions and experimental conditions.
• This effect persisted even when accounting for motor response time and manipulating prior expectations.

Conclusions:

• The visual system demonstrates positive hysteresis when interpreting complex, naturalistic scenes with changing semantic categories.
• Positive hysteresis acts as a stability-promoting mechanism, favoring perceptual history to ensure coherent interpretation of continuous visual input.
• These findings extend the understanding of perceptual hysteresis to dynamic, real-world visual environments.