Validity and Tests Developed on Reduced Redundancy, Language Components and Schema Theory

Abstract

This study reports the performance of 430 undergraduate and graduate students of English on standard C-Tests, the disclosed TOEFL, lexical knowledge test (LKT) and semantic schema-based cloze multiple choice item test (S-Test). Among the four tests, the first has been shown to be affected by their context. Two other versions were, therefore, developed on the standard C-Tests, i.e., Context Independent (CI)C-Test, and Context Dependent (CD)C-Tests from which the CIC-Test items were removed. Inspired by Kamimoto (1993), the well functioning items of standard C-Tests developed by Klein-Braley (1997) and their CD and CI versions were employed to study their underlying factors. The factorial analysis of the three C-Tests revealed two latent variables representing reduced redundancy and another unknown construct. The inclusion of the TOEFL, LKT and S-Test in the analysis to illuminate the nature of the second factor, however, resulted in the extraction of three latent variables. Since the standard C-Tests and CDC-Tests loaded the highest on the first factor along with other tests, it represented reduced redundancy. However, since the S-Test had its highest loading on the second factor upon which the other two measures of language proficiency loaded as well, it represented schema theory. The vocabulary subtest of the TOEFL loaded the highest on the third factor upon which the TOEFL itself and its structure subtest loaded as well revealing the componential nature of the TOEFL. The results are discussed and suggestions are made for future research.