A shortcut in language testing: Predicting the score for paper-based TOEFL based on one sub-score

Abstract

Using standardized tests such as paper-based TOEFL with three subtests for classroom assessment is restricted by the length of the test, which is usually longer than the class duration. Therefore, it is significant to be able to predict other subtests by conducting only one subtest. Therefore, the current study aimed to calculate prediction coefficients, enabling teachers to predict scores in paper-based TOEFL by conducting only one subtest. The data to create the prediction models were obtained from 2,030 scores of Institutional TOEFL, i.e. paper-based TOEFL without writing subtest. The prediction coefficient was calculated by using linear regression analysis. The result shows that the listening comprehension sub-score predicts the TOEFL score more accurately (MSE of 520) than other sub-scores (MSE of 553 and 587). The intercept for listening comprehension sub-score was 373.07, 357.14 for structure & written expression, and 364.19 for reading comprehension. In addition, the slope for each sub-score was 4.07, 5.96, and 4.63, respectively. Therefore, a listening test should be used in predicting the overall TOEFL scores for an accurate prediction.