Violating statistical structure impairs detection of deviant and incidental events.

Area of Science:

• Cognitive psychology
• Computational neuroscience
• Visual perception

Background:

• The brain learns statistical regularities in the environment to shape perception.
• A theory proposes percepts are biased toward expectations, with unexpected events increasing sensory gain.
• This mechanism balances accurate perception with reliable sensory estimates for updating internal models.

Purpose of the Study:

• To test the theory that perception is biased toward expected stimuli.
• To investigate how unexpected events impact performance and sensory gain.
• To understand how the brain updates its models when environmental statistical structures change.

Main Methods:

• Six experiments involved participants detecting visual stimuli.
• Stimuli were presented at the center or circumference of a circle.
• Bayesian changepoint modeling analyzed performance after disruptions in learned spatial or orientation regularities.

Main Results:

• Hit rates decreased for stimuli following a disruption of learned regularities (surprise).
• Performance recovery after a single change took multiple trials.
• Recovery became immediate when changes occurred more frequently, suggesting perceptual facilitation of the expected.

Conclusions:

• Perceptual systems broadly facilitate the expected, irrespective of stimulus latency.
• The study highlights a trade-off between perception and model updating when the environment changes.
• Further research is needed to understand how accurate environmental models are updated despite perception biases.