Abstract
Reading involves multiple eye movements and necessitates the integration of visual information from successive fixations. Here, we quantify the efficiency of trans-saccadic integration during foveal and peripheral word recognition by comparing performance for a human and an ideal observer. In Experiment 1, subjects were asked to identify random trigram letters presented at 13 possible positions on an invisible horizontal line at 0° or 10° eccentricity in lower visual field (horizontal distance from fixation dot: -6 to +6 letter slots). Trigrams were presented at a fixed size for different durations (62,125,250,500 ms followed by a post mask) while subjects steadily fixated a central dot. In Experiment 2, subjects were asked to identify single 7-letter words presented for 500 ms (followed by a post mask) at 0° or 10° eccentricity with lateral/horizontal positions identical to Experiment 1. First, a fixation dot appeared with '#' characters displayed at the word position. Subjects then pressed a button, which triggered the simultaneous appearance of a saccade-target dot and of the word. After execution of the saccade, subjects had to report the word. Eye movements were monitored to control fixation in Experiment 1 and to measure durations and positions of the two fixations in Experiment 2. Results for Experiment 1 show an inverted U-shape effect of the letter slot position on letter recognition performance. These visual-span curves (Legge et al, 2000) allow us to predict limitations in letter recognition performance for any fixation position and duration. Efficiency is then calculated using (1) human recognition performance from Experiment 2 and (2) recognition performance from an ideal observer based on visual span curves and a 7-letter word lexicon. Results show variability between subjects and suggest supplementary parameters to include in the model
Meeting abstract presented at VSS 2015