An EFL Readability Index

Abstract

This surdy explores readability and its relationship to the cloze passage performance of EFL students. Fifty reading passages were randomly selected from an American public library and made into 30-item cloze passages by deleting every 12th word. Thc subjects were 2298 EFL students from 18 university level institutions in Japan. Each student was randomly selected to take one of the 30-item cloze passages. Any differences between the cloze passages were therefore assumed to be due to other than sampling difierences. The result was a set of 50 cloze passages the means of which sewed as the dependent variable: EFL Difficulty. Each passage was then analyzed for two sets of independent variables chosen to investigate how well they predict the EFL Difficulty: the first set was made up of various first language readability indices (including the Flesch, Flesch-Kincaid, Fry, Gunning, Fog, and modified Gunning-Fog indices); the second set was made up of quantifiable linguistic characteristics of the passages (e.g., the percent of function words number of syllabuses per sentence, number of words per paragraph, frequencies of words in the passages, and nany others). Correlational, factor and multiple-regression analyses indicated that the first language readability indices were only weakly related to EFL Difficulty (23 to 30 percent). However, the analysis of linguistic characteristics indicated clear groupings among the variables. In addition, four of the linguistic characteristics (number of syllables per sentence, the average frequency of lexical items elsewhere in the passage, percent of words with seven or more letters, and percent of function words) when combined were more highly related to EFL Difficulty (55 percent). These results are discussed in terms of their implications for the development of an EFL readability index.