Abstract
Fixation durations have been shown to vary with the task definition and the onset of scene transients (Land & Hayhoe, 2001; Henderson, 2003). Recent studies have shown that fixation durations may be extended with global changes in the visual stimuli (Henderson & Pierce, 2008; Henderson & Smith, 2009; Pannasch et al., 2010). However these experiments were performed using static images that were interrupted by a noise image that was presented with varying durations. We wanted to study changes in fixation durations under a more natural setting, so we performed eye-tracking experiments while our subjects viewed movies on a computer monitor. Instead of using noise images, we used the numerous scene transitions in the movies to serve as global changes in the visual stimuli. Our analysis focused on fixation durations preceding the scene transition (control set), fixation durations after the scene transition (test set) and on-going fixations at the time of transition onset (cross-over set). We found the fixation durations for the test set to be significantly shorter than the fixation durations for the control set and the cross-over set (1-tailed 1-way ANOVA, p < 0.01). Subsequent multiple comparison tests showed that the distribution of fixation durations in the cross-over set was significantly larger than all other conditions. These results support the idea of Process Monitoring in eye movements (Henderson & Smith, 2009). Any global visual change appears to affect not only the length of ongoing cross-over fixations (immediate control), but also shortens the durations of the fixations immediately following that change (delayed control). Additional analyses revealed that this shortening effect lasted for about 220 milliseconds following a scene transition. These results show that fixations are under the influence of moment-to-moment visual and cognitive analysis.
This work was supported by the A*Star Human Factors Engineering Thematic Strategic Research Program.