Abstract
Eye-movements can be guided by two separate mechanisms: Bottom-up selection is a stimulus-driven mechanism that operates based on the uniqueness or saliency of visual features in a scene. Top-down selection is a goal-oriented mechanism that operates based on the context of the scene, or observer's intention. Although eye-movements are likely to be optimized to deal with dynamic natural scenes, studies of the influence of either top-down or bottom-up mechanisms on eye-movements have only used natural but static stimuli, or dynamic but artificial stimuli. To address this, the present study manipulated the goal-specificity of the task as a manipulation of top-down processing, and examined the influence on eye-movement behaviour while viewing dynamic natural scenes. In the first phase of the experiment, 40 observers were presented with short clips of a singles tennis match and were asked to rank nine items from the clips (e.g. players, ball, court lines, etc) from most to least attended (i.e. non-specific goal task). In the second phase, half the observers were asked to indicate which of the players won the point (i.e. specific goal task [SG]) in addition to performing the ranking task, while the rest performed the ranking task only (i.e. non-specific goal task [NSG]). Results revealed the significant impact of goal-specificity on the distribution of several eye-movement parameters such as saccadic amplitudes and fixation durations. Most importantly, under the SG task, observers showed shorter saccadic amplitudes and shorter fixation durations, especially for those parameters recorded spatiotemporally around the ‘ball events’ (i.e. serve, hit, bounce and net); this suggests that the SG task might facilitate finer gaze control. We conclude that top-down selection can override bottom-up selection in the guidance of eye-movements even in the dynamic natural scenes which contain highly salient visual features such as motion.
EPSRC grant (EP/F069421/1).