The Interplay of Visual Attention and Saccade Planning in Active Vision.

摘要

Vision is a highly active process. When we view the world, we do not hold our eyes still, but constantly move them around in order to view the object or area of interest with the fovea (the region of the retina with the highest acuity). Saccades are the step-like movements that we most often employ for this purpose. In addition, our attention is constantly being covertly attracted or directed to points of interest. Combining these different aspects of viewing: visual processing, the orienting of attention, and eye movements can be referred to as 'active vision'.;Most work on active vision or attention and saccades has concentrated on performance improvements preceding saccades, but relatively little is yet known about how attention affects later stages of saccade planning. That is the focus of this thesis. First we look at the temporal dynamics of the scaling of attention and what influence that attention scale exerts on the decision to make a saccade. We are able to infer the attention scale during individual trials from their saccade latencies. We find that the scale of attention changes very rapidly, and faster than previously reported. The remainder of the thesis concentrates on the effects of attention scale and locus on post-saccade adaptive processes: how the success of the current plan influences learning.;Saccades maintain their accuracy through an adaptive process, slowly to compensate for muscle weakening, or rapidly in a lab setting using intra-saccadic steps. Little is known about how covert attention interacts with this process. The second study of this thesis looks at how the scale of attention can affect the magnitude of saccade adaptation. We use a novel paradigm in which the intrasaccadic steps change from trial-to-trial so that over many trials the displacement produces a sine wave pattern. We find that when attending to larger targets, there is proportionally greater adaptation than when attending to smaller targets. Finally, we demonstrate that the locus of attention at the end of a saccade contributes to the error signal for saccade adaptation. Instead of intra-saccadically moving the target in order to induce saccade adaptation, we present a distractor briefly after the saccade on the near side or far side of the target. By drawing attention away from the saccade target immediately after the saccade, the distractor is able to induce saccade adaptation. The magnitude of the saccade adaptation depends on the novelty of the distractor.;These experiments highlight the interplay between attention and saccades. Using novel paradigms, we show that the locus of attention can induce saccade adaptation, and that the scale of attention influences both the magnitude of saccade adaptation as well as the decision to move. While conventionally, saccade experiments are performed using very small stimuli, we see that using larger stimuli can greatly change saccade performance. Use of larger, more complicated stimuli as compared to simple spot targets is a step closer to natural viewing and very important to our understanding of active vision.