Research News


The Best Way to Evaluate ESL Texts for Reading Effort

image picture
Image by HQuality/Shutterstock

A researcher at the University of Tsukuba in Japan shows that readability formulas incorporating cognitively based features predict English as a Second Language reading effort better than traditional formulas

Tsukuba, Japan—Reading skills develop and improve when texts are matched to the reader's current ability in a way that is both challenging and achievable. Methods to measure text readability are thus an important issue for a wide range of reading applications, such as education, production of reading material, and publishing. Numerous tools to assess readability have been developed that most often rely on classical formulas, such as Flesch Reading Ease (FRE), which considers word and sentence length. "Most formulas focus on reading comprehension but not processing," says Professor Shingo Nahatame, who took a new approach in his research undertaken at the University of Tsukuba in Japan. "The focus of this research was different. I assessed traditional and newer formulas in their ability to predict reading effort in a second, non-native language." The research findings were published in the scientific journal Language Learning.

Newer formulas for determining readability consider the cognitive aspects of reading, in addition to the traditional measures. In particular, new formulas take into account the lexical sophistication, syntactic complexity, and cohesion of the reading material—how well different parts of the text are tied together. To determine which formulas were best for assessing English as Second Language (ESL) reading effort, Nahatame first calculated the readability scores for numerous reading materials based on both traditional and new formulas. He then measured the eye movements of Japanese college students as they read the material. He also analyzed eye-movement data from Dutch ESL speakers that were extracted from an open eye-tracking database. Eye movement data provided four direct measures of reading difficulty, rather than inferred measures such as reading duration.

These measures were: (1) The time spent fixating on words between eye movements, (2) the length of the movements, (3) the frequency of skipping words, and (4) the frequency of regressive eye movements back to previously read text. In general, when text is difficult to understand, eye fixations are longer, distance of eye movements are shorter, fewer words are skipped, and there are more regressions.

He found that when reading material was very short, none of the traditional formulas predicted any of the eye movements measures, and a newer formula only predicted fixation duration. When he examined longer reading material, even though traditional methods did predict fixation, the non-traditional formulas also predicted skipping rates and the magnitude of eye movements. "Overall, newer models that consider the cognitive aspects of reading are more accurate than the traditional models at assessing reading effort in ESL," says Nahatame. "However, no single formula was the best on all measures on both tests." Thus, when evaluating text for reading effort, a combination of formulas is likely to be the best strategy.

This research was supported by JSPS KAKENHI Grant Number 20K00827.

Original Paper

The article, "Text Readability and Processing Effort in Second Language Reading: A Computational and Eye-Tracking Investigation," was published in Language Learning at DOI: http:doi.org/10.1111/lang.12455


Assistant Professor NAHATAME Shingo
Faculty of Human Sciences, University of Tsukuba

Related Link

Faculty of Human Sciences

Inquiry for collaboration (Flintbox)

Celebrating the 151st{50th Anniversary of the University of Tsukuba
Celebrating the 151st{50th Anniversary of the University of Tsukuba