Abstract
Eye movement (EM) analyses for static images are well defined, but less so for dynamic stimuli. There are limits in applying static measures, such as fixation duration, to dynamic stimuli, that prevent the appropriate characterization of eye movements over space and time. Consequently, moving stimuli are often analyzed using ‘swarm analysis’ or dynamic heat maps to describe EM. Although these allow for qualitative representations of EM, it is difficult to quantify similarities between observers. Previous researchers have used the separate variance in X and Y gaze position to describe EM variability, however, in clinical research, two measures of EM stability combine X and Y gaze position: the bivariate contour ellipse area (BCEA), and the within-isoline area. The BCEA calculates the highest density of eye movements using a Gaussian ellipse that encompasses 68% of the fixation data, thus assuming normality in EM. The within-isoline area does not assume normality. It is based on the probability density estimate of EM, with a level of density chosen (e.g., 68%), so that 68% of the data points have a higher density estimate than this level. Here we apply these metrics to EM generated while watching either an intact version of an emotionally salient movie, or a version where the movie’s scenes were scrambled in sequence. EM were recorded using a 60Hz binocular tracker (Mirametrix S2), with sample gaze positions analyzed in MATLAB using two algorithms. We found that both algorithms show EM are more variable during the scrambled movie in comparison to the intact movie. However, the BCEA shows greater variance in EM within and across both conditions in comparison to the within-isoline area. We attribute this difference due to the assumption of normality in the BCEA calculation, and recommend the within-isoline area when describing group variance of EM in dynamic stimuli.
Meeting abstract presented at VSS 2015