Defending Yarbus: eye movements reveal observers' task.

Area of Science:

• Cognitive Science
• Computer Vision
• Neuroscience

Background:

• Alfred Yarbus's research suggested eye movements reflect task demands.
• Previous studies yielded conflicting results on decoding observer tasks from eye movements.

Purpose of the Study:

• To systematically investigate the informativeness of eye movements for task and mental state decoding.
• To re-evaluate previous findings and extend Yarbus's original experiments.

Main Methods:

• Reanalysis of existing eye-movement data using a Boosting classifier.
• Collection of new eye-movement data from observers viewing natural scenes under specific questions.
• Analysis of aggregate eye-movement features for task decoding.

Main Results:

• Task decoding was achieved significantly above chance (34.12% vs. 25% chance) in the reanalyzed data.
• Task decoding was also successful in the extended Yarbus experiment (24.21% vs. 14.29% chance).
• Results support the hypothesis that eye movements can reveal observer tasks.

Conclusions:

• Yarbus's hypothesis regarding task-modulated eye movements is supported by the findings.
• Eye movements provide informative signals for decoding tasks and mental states.
• The study encourages further research in computational and experimental eye-movement analysis.