Oculo-motor patterns induced by reading in peripheral vision

Abstract

There is clear evidence that slower reading in peripheral vision results from shrinkage of the visual span - the number of letters recognized with high accuracy without moving the eyes. Based on this evidence, an ideal-observer model, Mr. Chips (Legge et al., 2002), has been used to simulate saccade planning in reading. This model shows a strong relationship between the size of the visual span and saccade length. One prediction of this model is that reading speed in peripheral vision should be correlated with the saccade length (or more precisely with the horizontal component's length of saccades). This prediction is based on the implicit assumption that regression saccades do not occur too often. We have investigated this issue by measuring eye movements of 34 patients with central field loss (induced by age-related macular degeneration). Patients had to read aloud 14 French sentences displayed in succession (each sentence was displayed on one line). Character size was 3× the individual ETDRS acuity. Each patient had an absolute macular scotoma covering the fovea as assessed by MP1 microperimetry. Ocular data were collected with an Eyelink II eyetracker (500 Hz). The horizontal distribution of fixations was analyzed for each sentence with kernel density estimates. In addition, we assessed whether these distributions contained regions with statistically significant curvature – i.e., regions with a high density of fixations (called clusters hereafter). Results show a high variability of density estimates both within- and between patients: some sentences exhibit very homogeneous distributions while others show clear density peaks. The main finding is that reading speed is significantly slower for sentences that contain fixation clusters compared to sentences without clusters. This relationship between reading speed and the presence of fixation clusters has to our knowledge never been reported and should be taken into account to understand eccentric reading.