Testing temporal integration of feature probability distributions using role-reversal effects in visual search

Area of Science:

• Cognitive Neuroscience
• Visual Perception
• Computational Neuroscience

Background:

• The human visual system processes complex scenes by encoding statistical properties, including feature probability distributions.
• It remains unclear if the brain constructs probabilistic models of dynamic visual input using these statistics.

Purpose of the Study:

• To investigate whether the brain utilizes statistical properties of visual input to build probabilistic models.
• To examine how observers temporally integrate orientation distributions from sequential visual search trials.

Main Methods:

• Participants performed sequential odd-one-out visual search tasks.
• Distractor orientations were sampled from two Gaussian distributions with manipulated variances.
• A test trial with switched target/distractor roles assessed internal representations via response times.

Main Results:

• Search times on test trials were primarily influenced by the most recent learning trial, showing a strong recency effect.
• This recency effect suggests a reliance on the latest stimulus when input reliability is low.
• Temporal integration was observed, indicating the visual system's preference for stable environmental statistics.

Conclusions:

• The visual system prioritizes recent stimuli when dealing with unreliable or changing visual input.
• Sensory history is utilized when environmental statistical properties are stable.
• This suggests a flexible strategy in probabilistic inference based on environmental reliability.