Abstract
Two important aspects of eye-movement control during scene viewing are where fixations tend to be directed (fixation position), and how long they typically remain there (fixation duration). While substantial behavioral and computational research has been devoted to fixation positions in scenes, relatively little is known about the mechanisms that control fixation durations during scene viewing. We have recently investigated the degree to which fixation durations are under direct moment-to-moment control of the current visual scene using a scene onset delay paradigm. During saccades prior to specified critical fixations, the scene was removed from view so that when the eyes landed in fixation, the scene was absent. Following a pre-defined delay (0, 300, 400, 600, 800 ms, or infinite delay), the scene reappeared. Two populations of fixation durations were observed: A first (early) population of fixations that terminated during scene absence independently of the delay, and a second (late) population of fixations that increased in duration as delay increased. Here, we propose a random timing model with inhibition and two-stage saccade programming to account for the observed fixation duration patterns. Model assumptions are consistent with current evidence concerning basic oculomotor control. First, an autonomous (i.e., random) saccade timer generates variations in fixation durations. Second, we assume that difficulties at the level of visual and cognitive processing can inhibit saccade initiation, essentially leading to longer fixation durations. Third, saccade programming is completed in two stages: an initial, labile stage that is subject to cancellation, and an ensuing, non-labile stage. The model was tested on the scene onset delay data. Simulations qualitatively reproduced the mixed-control pattern of fixation durations observed in the empirical data. We conclude that fixation durations, to a certain degree, reflect perceptual and cognitive activity in scene viewing; computational model simulations contribute to our understanding of the underlying processes.