Human visual search does not maximize the post-saccadic probability of identifying targets

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Vision

Background:

• Previous research suggested eye movements in visual search aim to minimize saccades (rapid eye movements).
• An optimal Bayesian strategy involves using the entire retina for information gathering, not just targeting likely locations.

Purpose of the Study:

• To investigate whether human observers minimize saccades during a simple visual search task.
• To determine if eye movement strategies adapt to changes in visual search parameters like token spacing and size.

Main Methods:

• A simple visual search task involving three tokens was presented to human observers.
• Experiments manipulated token spacing and token size to alter the optimal eye movement strategy.
• Observed saccade planning and strategy selection were compared against theoretical optimal strategies.

Main Results:

• Human observers did not minimize the expected number of saccades in the visual search task.
• Observed strategies remained constant despite variations in token spacing and size, unlike optimal strategies.
• Human performance was qualitatively and quantitatively suboptimal compared to ideal Bayesian models.

Conclusions:

• Human visual search strategy does not align with the hypothesis of minimizing saccades.
• Observers exhibit inflexible eye movement planning, failing to adapt to task-specific optimal strategies.
• Current models of eye movement control may need revision to account for human limitations in visual search.