Measures of eye movements

When examining global indexes of eye movements in reading, reliance on a single measure can be misleading. For example, a decrease in average fixation duration for a text could be offset by an increased number of fixations³². The total number of fixations, on the other hand, might be a byproduct of perceptual span, but is also influenced by regressions and re-reading. It is thus important to consider multiple indexes. Here, we relied on (1) average fixation duration, (2) number of fixations, and (3) average amplitude of the saccades. Average fixation duration is assumed to capture the time, and thus the effort, needed to decode and process the orthographic input. Longer durations are assumed to signal less readable texts. For texts, number of fixations and average amplitude of the saccades map onto the construct of perceptual span (e.g.^25,32), that is the amount of information that can be processed during fixation. When this amount decreases, the saccades will cover a smaller extension, and more fixations would thus become necessary to finish reading. Further, in our work, number of fixations captures also the potential need of a second reading for some part of the text. Increased number of fixations and reduced saccade amplitude would thus be both associated to reduced readability.

To extract these three measures, for each webpage all the parts of the main text were identified as areas of interest (AIs). This segmentation excluded pictures and any other element that was not part of the main text (e.g., links, advertisements). Eye movement analysis was limited to the phenomena detected within the AIs, in order to better capture eye movements related to actual reading.

Within each page, blink-related artifacts were removed, by excluding fixations recorded immediately before or after the blinks and lasting less than 100 ms. For each participant and each page, durations of all the fixations (within the text area) were averaged together, obtaining average fixation duration. This measure was used as an index of the effort needed to decode and process the orthographic input provided by the texts.

The same fixations used to compute average fixation durations were used for this measure. The total number of fixations was counted within each webpage and each participant. This measure captures both perceptual span (together with the average amplitude of the saccades, described below), as well as regressions or second-pass reading.

For each page, saccades containing blinks were removed. Further, only saccades having their starting and end points within the textual AIs were retained, in order to focus on reading related movements. The amplitudes of these selected saccades were averaged within each webpage and within each participant. Together with number of fixations, this measure was mainly considered as an index of perceptual span.