Abstract
Purpose:
Previous studies have shown that readers often have small errors in binocular fixation during reading, This study investigated the amount of disparity that can be tolerated before reading becomes difficult. Fixation error was manipulated by shifting disparity during saccades, using an added noise paradigm.
Methods:
A dual Purkinje image eye tracker was used to track the eye movements of a convenience sample of 10 subjects with self-reported normal reading ability. Reading rate was measured while reading aloud randomly selected sentences from the MNRead database in two conditions (horizontal or vertical disparity noise) with 45 trials each. A horizontal saccade contingent position error added noise paradigm was used so that the sentence disparity shifted a certain amount during each saccade, varying from 0 to 192 arc minutes. Reading time was plotted against the standard deviation of the errors added with each saccade and a two line least squares fit was used to determine the amount of noise that just begins to increase reading time over baseline (“critical disparity jitter”). The presence of a vergence response by the subjects was verified by analysis of the eye movement record of both eyes in response to the disparity target to ensure the eyes followed the target properly and suppression was not an issue.
Results:
Horizontal critical disparity jitter ranged from 0.038 to 1.136 degrees (mean=0.534, SD=0.35). Vertical critical disparity jitter ranged from 0.02 to 0.52 degrees (mean= 0.15, SD=0.10). This corresponds to a disparity jitter of 2.67 character widths horizontally and 0.375 lines vertically where the subjects still read at baseline rate.
Conclusions:
Relatively large amounts of binocular disparity errors are tolerable horizontally, even if the disparity is large enough to produce double vision. Disparity is less tolerable vertically.
Keywords: 672 reading •
522 eye movements •
434 binocular vision/stereopsis