Eye tracking and paragraph reading: Learning about return-sweep performance


University of Utrecht


From a perceptual point of view, a paragraph is a very complex and tangled unit of information. They contain a multitude of shapes that can’t be interpreted all at once. Navigation within these blocks of text is thus essential.

For my masters thesis in Utrecht University, I joined Jessica Heeman’s eye-tracking laboratory to learn more about the challenges of paragraph reading. I focused on return sweep performance, which is the rapid eye movements that happen between lines of text.

Since early times, type workers have used different tools to help readers with such navigation. In the 15th century, a common practice was to add the first word of the next page at the bottom of the current one. Below is Erotemata (Sébastien Gryphe, 1544), displayed at the Library of Lyon (Image courtesy of Peter Bilak.)

Additions to improve performance

Taking inspiration from the repetitions 15th-century type workers used to aid readers, I delved into which visual cues would be most useful for reader's ease of navigation within the paragraph. After a review of recent legibility literature on line navigation, I selected three additions that I expected to have the greatest influence on readability.

Recent research suggests readers seem to ignore repetitions if the words are coherent with the text (Ehrlich and Rayner, 1981; Drieghe, et al., 2010; Rayner, et al., 2011). These could then hypothetically help readers locate the start of the new line, for which two conditions were based in repeating the last word of each line before and within the next line. The third condition aimed to increase line differentiation by adding diamonds before each line (see below). A control condition was also added to compare against.

Twenty passages from The Little Prince were used to generate the stimuli of the experiment (by Antoine de Saint-Exupéry, originally published in 1943. Fragments were based on Richard Mathews' English translation of the book). They were matched in length and complexity, and were shown in either of the conditions randomly and in a random order.

Below are four adapted examples of the stimuli. In the experiment, all text was justified to the right and in Helvetica, using optimal text size and line separation to reduce head movements. These decisions were based on a pilot study conducted prior to the experiment.


Legibility was operationalized using four parameters: Speed of reading; amount of leftward saccades, that is, eye movements to re-read a passage; amount of fixations and correct answers to comprehension questions about each text.

All eye movement parameters were measured using an eye tracking device EyeLink 1000. The device allowed to collect one thousand recordings per second while participants were reading the passages. Data from twenty participants was used.


Results showed that, out of the three additions studied, diamonds before each line outperforms the others when it comes to reading speed, re-reading eye movements and amount of fixations. This difference is only significant against the other two text additions (outside and within the text), and scores from the control always remained between text and graphic additions for the three measurements of legibility considered.

These results align with previous research, which supports line differentiation is key to improve readability, as parafoveal information is known to guide return sweeps. Parafoveal information is the stimuli generated in the periphery of the retina. This stimuli is lower in density, and has less sharpness than the information provided by the fovea, the focal area of the retina.

The information density the two text conditions provide was probably impractical, as the additions are probably only processed after each return sweep, hindering performance. When landing on the new line, participants were probably confused after finding the same word. Participants weren't able to accustom to the text additions, and the effects of training were not considered.

Below is a representation of the visual acuity of a reader when focusing on the word “will”, at the end of the second line. Based on Tan, Weimin & Yan, Bo. (2017). The further from the foveal point, the less acuity.  The saliency of each line is wildly different between both conditions, a potential explanation to these results.

This is a reduced version of the original research project. You can read the full manuscript via the Univeristy of Utrecht's thesis repository.

Ehrlich, S. F., & Rayner, K. (1981). Contextual effects on word perception and eye movements during reading. Journal of Verbal Learning & Verbal Behavior, 20(6), 641–655. https://doi.org/10.1016/S0022-5371(81)90220-6

Rayner, K., Slattery, T. J., Drieghe, D., & Liversedge, S. P. (2011). Eye movements and word skipping during reading: Effects of word length and predictability. Journal of Experimental Psychology: Human Perception and Performance, 37(2), 514–528. https://doi.org/10.1037/a0020990

Saint-Exupery, A. de, De Saint-exupery, A., & Howard, R. (2020). The Little Prince: Translated by Richard Howard. Van Haren Publishing.


Acknowledgements: Dr Jessica Leeman, Peter Biľak

See more