Quantitative analysis of saccadic search strategy

Abstract

This thesis deals with the quantitative analysis of saccadic search strategy. The goal of the research presented was twofold: 1) to quantify overall characteristics of fixation location and saccade direction, and 2) to identify search strategies, with the use of a quantitative description of eye movement parameters. Chapter 2 provides a method to quantify a general property of fixation locations. We proposed a quantitative measure based on Voronoi diagrams for the characterization of the uniformity of fixation density. This measure may be thought of as indicating the clustering of fixations. We showed that during a visual search task, a structured (natural) background leads to higher clustering of fixations compared to a homogeneous background. In addition, in natural stimuli, a search task leads to higher clustering of fixations than the instruction to freely view the stimuli. Chapter 3 provides a method to identify the overall field of saccade directions in the viewing area. We extended the Voronoi method of chapter 2 so that it became possible to create vector maps. These maps indicate the preferred saccade direction for each position in the viewing area. Several measures of these vector maps were used to quantify the influence of observer-dependent and stimulus-dependent factors on saccade direction in a search task with natural scenes. The results showed that the influence of stimulus-dependent factors appeared to be larger than the influence of observer-dependent factors. In chapter 4 we showed that the border of the search area played a role in the search strategy. In a search experiment in differently shaped areas we measured that search performance was poorer near the search area luminance edges. Fixation density, however, was higher in the edge region, and saccade direction was mainly along the edges of the search areas. In a target visibility experiment we established that the visibility of targets near a luminance edge is less than the visibility of targets against a homogeneous background. We interpreted this as the visual span being smaller near luminance edges than in homogeneous backgrounds. We concluded that search area edges affected eye movement strategy both at sensory level and cognitive level. Search area edges hampered target perception, so that fixation density had to be higher in the edge regions in order to achieve similar search performance as in the center regions. In chapter 5 we investigated the effect of a priori knowledge about target conspicuity on saccade amplitude and fixation duration. The visual system may use fixation duration and saccadic amplitude as optimizers for visual search performance. We expected to find (nearly) optimal settings for saccade amplitude and fixation duration in a search task with known target conspicuity, and sub-optimal settings in a search task with unknown target conspicuity. However, we found that saccade amplitude decreased and fixation duration increased gradually as a function of the serial order of saccades and fixations in search experiments with known as well as unknown target conspicuity. These changes are the mark of a coarse-to-fine search strategy, in which a stimulus is at first scanned quickly, and progressively better scrutinized to find the target. We hypothesized an intrinsic coarse-to-fine strategy for visual search that is even used when such a strategy is suboptimal.