Measuring Readability for Japanese Learners of English

This paper describes the relative effectiveness of seven variables of three categories in predicting the readability of the EFL texts used in the Japanese context. The factors examined in our research were (1) word difficulty and (2) idiom difficulty, in addition to the commonly used variables, sentence length (SL) and word length (WL). In the analysis of word difficulty three measures were considered: (1) JACET 8000 (JACET), (2) the British National Corpus (BNC), and (3) Textbook-based Word Difficulty (TWD). As measures of idiom difficulty two factors were considered: (1) Akao-based Idiom Difficulty (AID) and (2) Textbook-based Idiom Difficulty (TID). The textbook-based word difficulty and textbook-based idiom difficulty are original measures being developed in the present project. Six multiple regression analyses computed using six combinations of four kinds of variables, revealed the following points. First, the combination of variables, SL + WL + TWD + TID, proved to be the most efficient predictor of readability for the EFL sample textbook materials. Second, the regression equation that best explains the total textbook corpus was: Year level = 0.0995*Words/S + 0.4302*Syllab/W + 0.9799*WordDiff/W + 0.0633*IdiomDiff/S + 0.2815.