ComprehensionWorkshop

Investigating sentence reading in young adults with and without university education: An eye-tracking study

Authors:
Thul, Rüdiger, ruediger.thul@nottingham.ac.uk, University of Nottingham
Conklin, Kathy, k.conklin@nottingham.ac.uk, University of Nottingham

Keywords: reading authentic texts, young adults, non-WEIRD participants, generalized linear modelling

Abstract:

While significant work has been done to understand reading in children and specialized adult populations, we have much less of an understanding of the potential variability in healthy adult readers across a broad cross-section of society for whom decoding and reading is automatized and where the primary aim of reading is to understand written texts and gain new information.

The current study addresses this gap by monitoring eye movements during the reading of authentic texts by unimpaired, matched native English-speaking populations with (n = 13; mean age = 18.6) and without (n = 10; mean age = 21) university education. Participants read 10 BBC articles (total words = 7,444 and 163,212 data points). After reading an article, participants rated it for: informativeness, ease of reading, being well-written and being enjoyable.

There was a strong correlation between an article being perceived as easy to read and as being well written amongst participants without a university education (ρ = 0.67). However, this relationship was less evident in university student participants (ρ = 0.28). We employed generalized linear mixed models for first fixation duration and total reading time and found that both reading measures were best captured by a distribution from the Box-Cox family. Compared to participants with university education, participants without university education had longer and more variable first fixation durations (β~1~ = 0.094, t = 10.1; β~2~ = 0.061, t=10.82) and longer, but less variable total reading time (β~1~ = 0.118, t = 9.88; β~2~ = -0.046, t = -2.45). Both group’s first fixation durations and total reading time decreased with increasing Zipf values (β~3~ = -0.02, t = -16.06; β~3~ = -0.056, t = -32.8), and so did variability in reading time. For articles that participants rated as challenging and less enjoyable, there was a greater processing costs for low frequency words for participants without a university education.

Our work demonstrates that it is critical to move beyond datasets from Western, Educated, Industrialized, Rich, and Democratic (WEIRD) populations in reading research. Crucially, we show that less-WEIRD data (young people not pursuing a university education) is more heterogenous. Finally, gaining a greater understanding of how reading changes in an authentic context and between groups of readers may help writers/authors to make their texts more accessible for a wider audience.