Jerry Greenfield

Classic readability formulas in an EFL Context - Are they valid for Japanese Speakers?

(Dissertation Abstract)

Classic readability formulas have been widely used to match English texts to native English speaking readers, for whom the formulas have been extensively validated. They are also being used in ESL and EFL contexts, for which they have not been conclusively validated. Recent research on second language reading, together with consideration of the nature of the formulas and assumptions on which they rest, raise questions about their validity for EFL readers.

This project investigates the validity of the Flesch, Flesch-Kincaid, Bormuth, Coleman-Liau, and New Dale-Chall formulas for predicting the readability of English texts for Japanese college students. Validity is tested in multiple ways, using a new criterion established by cloze testing of 200 Japanese students over 31 passages assembled by Bormuth. The observed difficulty of the criterion passages is compared with their difficulty predicted by each of the classic formulas, finding Pearson correlations ranging from .691 to .861. Using the models of the classic formulas, new regressions are made against the new criterion, and their predictions of difficulty are compared with predictions by each of the original formulas. For Coleman-Liau, Bormuth, and Dale-Chall models the recalculated formulas were found to be significantly more accurate than the original formulas. In addition, a new Miyazaki EFL Readability Index was constructed using only two easily measured text variables. Producing scores on a 100-point scale, the formula has an adjusted coefficient of determination of .726, superior to all but the recalculated Bormuth formula. Finally, the relationship of observed difficulty to readers’ TOEFL scores was explored and found to be too weak to support indexing readability to TOEFL proficiency of this group.

On the strength of these findings, the classic formulas are probably valid for use with Japanese EFL students, although they require interpretation. Increases in accuracy found for recalculated formulas, while statistically significant, may not be sufficient to replace formulas conveniently available for computer measurement. However, the new Miyazaki Index offers an easy to use alternative that also appears to be highly valid.

The dissertation is available from University Microfilms International as UMI document number 9938670 (Year 1999).