When preview information starts to matter: Development of the perceptual span in German beginning readers

Abstract

How is reading development reflected in eye-movement measures? How does the perceptual span change during the initial years of reading instruction? Does parafoveal processing require competence in basic word-decoding processes? We report data from the first cross-sectional measurement of the perceptual span of German beginning readers (n = 139), collected in the context of the large longitudinal PIER (Potsdamer Intrapersonale Entwicklungsrisiken/Potsdam study of intra-personal developmental risk factors) study of intrapersonal developmental risk factors. Using the moving-window paradigm, eye movements of three groups of students (Grades 1–3) were measured with gaze-contingent presentation of a variable amount of text around fixation. Reading rate increased from Grades 1–3, with smaller increases for higher grades. Perceptual-span results showed the expected main effects of grade and window size: fixation durations and refixation probability decreased with grade and window size, whereas reading rate and saccade length increased. Critically, for reading rate, first-fixation duration, saccade length and refixation probability, there were significant interactions of grade and window size that were mainly based on the contrast between Grades 3 and 2 rather than Grades 2 and 1. Taken together, development of the perceptual span only really takes off between Grades 2 and 3, suggesting that efficient parafoveal processing presupposes that basic processes of reading have been mastered.

Keywords :

• Eye movements
• German
• Moving window
• Perceptual span
• Reading development

We thank Petra Schienmann for invaluable help with data collection.

No potential conflict of interest was reported by the authors.

This work was supported by the German Research Foundation (DFG) [grant number GK-1668/1] as part of the Research Training Group “Intrapersonal developmental risk factors in childhood and adolescence: A longitudinal perspective”.

Notes

¹ Since the focus of the longitudinal PIER study is on individual differences, all participants were presented with the same combinations of window sizes and sentences. This potentially sacrifices generalisability across materials in order to maximise between-subjects comparability by treating all participants with the same conditions, thereby supporting correlational analyses of individual differences in the context of the PIER study (compare Häikiö et al., 2009; Joseph et al., 2013; or Blythe & Joseph, 2011 on these concerns).

² A relatively large number of age-appropriate sentences were used in an attempt to keep processing difficulty constant between groups. Additionally, all participants read first-grader sentences. Rather than running separate analyses, we decided to model data of all sentences (age-appropriate and first-grade), and to include sentence type as a predictor, thereby improving statistical power with respect to the number of items and reducing specificity in the sentence-to-window-size assignment. More specifically, the inclusion of first-grader sentences improved estimability of item variance.

³ Refresh-rate differences did not significantly affect any of the reading measures in the reported mixed linear models.

⁴ Note that polynomial contrasts with window size treated as a categorical variable are used here merely to offer a description of the data pattern for the window sizes tested. A more realistic model with more easily interpretable parameter estimates would assume that window-size effects reach an asymptote for large enough windows; however, the fitting of such a model, e.g. of the form y = β ₁ + (β ₁ − β ₂)exp(−β ₃ x), would require non-linear mixed-effects models, which are harder to handle and less well developed (see Discussion).