Purchase this article with an account.
Rebecca M. Foerster, Werner X. Schneider; The influence of prediction violations on eye movement patterns in a LTM-driven multi-step sensorimotor task. Journal of Vision 2014;14(10):93. doi: https://doi.org/10.1167/14.10.93.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Prior studies revealed that prediction-violating stimuli are more often and longer fixated than prediction-confirming stimuli. This effect has been found in visual-search, change-detection, and memory-recall tasks. It is interpreted as prioritized processing of prediction-violating (surprising) stimuli. For preparing an action, action-relevant features have to be predicted and tested. Irrelevant features are usually completely ignored. Until now, it has not been tested how prediction-violating features that are either action-relevant or action-irrelevant influence gaze patterns in a sensorimotor task. We manipulated predictability and action-relevance in a computerized version of the number-connection task. Participants clicked on numbered shapes in ascending order. 60 trials were performed with the same spatial arrangement of 8 numbered shapes. In the consecutive 20 trials, the locations of two action-irrelevant features (shapes surrounding number 3 and 6) were exchanged in one group while the locations of two action-relevant features (numbers 3 and 6) were exchanged in another group. In a third group both changes appeared and no changes were executed in a control group. In 20 final trials, participants had to work again on the original configuration. Results revealed worse clicking performance, more fixations, and shorter fixation durations after an action-relevant change (switched numbers) only. The effects lasted for several trials. Micro-analyses revealed that the effects were completely due to an enhanced number of searching fixations but not guiding fixations, especially while searching for number three. We conclude that predictions for multi-step actions are formed only for action-relevant features and tested in each run. In addition, elongated processing of prediction-violating stimuli is not obligatory, but task-dependent.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only