Abstract
Introduction: Strategic deployment of eye movements is critical for maximizing perceptual performance in many tasks, ranging from visual search (Najemnik & Geisler, 2005) to face recognition (Peterson & Eckstein, 2011). Little is known about how these eye movement strategies are learned. Here, we assessed the human ability to learn optimal eye movement strategies with unusual face stimuli. Methods: We first measured observers’ preferred point of fixation for face recognition by having observers identify briefly presented faces. Observers then ran 1600 trials identifying four morphed faces which, unknown to the participants, contained all discriminating information within the mouth area. This study interleaved free eye movement trials and forced fixation trials where observers were required to maintain fixation at their preferred saccadic landing point. Next, we measured a visibility map by having observers identify the morphed faces while maintaining fixation at 5 specific locations. Finally, observers completed 500 additional trials in which free eye movement and forced fixation trials were interleaved. Results: Observers were classified into two groups: learners (who showed a large and significant increase in performance in both fixed and free conditions) and non-learners. Learners, unlike non-learners, also exhibited a significant downward shift in saccade landing location across trials; however, performance improvements were similar for the free and forced fixation conditions. Visibility measurements showed that while eye movements migrated downward to the nose tip, they did not reach the optimal fixation point (center of the mouth). Subsequent learning sessions after exposure to various points of fixation resulted in both initial learners and non-learners executing optimized eye movements that maximized performance. We suggest that humans’ initial failure to learn optimal fixation strategies with unusual faces arises from a difficulty to break away from over-practiced eye movement strategies that are optimal across an ensemble of regular faces but not for unusual faces.
Meeting abstract presented at VSS 2012