Abstract
Humans move their eyes via saccades several times per second to informative locations of the environment while performing natural tasks. We investigated how saccades in such a task change during learning and automatization and whether eye movement patterns after extensive training are similar for visible and invisible objects. To that aim, we applied a bimanual, high-speed motor task requiring grasping, moving, rotating, and putting down objects. Eye movements were recorded during the training phase and also compared under visual illumination and in the dark after training. With increasing automatization of the task, the eye-hand leading time became smaller and fewer fixations were made, while fixation rates remained constant. The comparison between normal vision and darkness revealed lower fixation rates and longer fixation durations in the dark, while eye-hand coordination and scanpaths were very similar. Results indicate that a distinct saccadic pattern is learned and transferred to long-term memory during automatization. After the training period, sequences of saccades are retrieved from long-term memory. These preprogrammed saccades are executed even in the dark. Interestingly, scanpaths similarity between normal illumination and darkness conditions was not related to task performance. Possibly, saccadic eye movements have no function, when no visual feedback is available. However, the inhibition of automatized eye movements might have an additional cost. Deductively, results point to a shift from sensory-based to long-term memory based visual selection during learning and automatization.
This research was supported by grants from the Cluster of Excellence Cognitive Interaction Technology (CITEC).