The effect of the computerization of the TOEFL on the English language proficiency testing of international students at the University of Mississippi

The Test of English as a Foreign Language (TOEFL) currently has two different forms, the Paper-based TOEFL (PBT) and the Computer-based TOEFL (CBT). The CBT represents advances in information technology that have allowed test publishers to incorporate computer-adaptive testing (CAT) techniques and to employ question types that are unavailable for paper-based testing. Computerization has lead to greater costs for students as well as potential for greater measurement accuracy. The purpose of this dissertation was to determine the effect of the computerization of the TOEFL on the measurement of ESL proficiency of incoming international students at The University of Mississippi (UM) by examining the correlation between the University's placement test, comprised of the Michigan Test of English Language Proficiency and the Listening Comprehension Test (MTELP/LCT), and the two different versions of the TOEFL, the CBT and the PBT.; This was a post hoc study intended to correlate already existing data for the purpose of examining them in new ways. The participants in the study were those who had submitted TOEFL scores as part of their applications to graduate and undergraduate programs at UM and who took the MTELP/LCT upon arrival on campus in August 2001. Fifty-three submitted PBT scores; 39 submitted CBT scores. Pearson product-moment correlation analysis revealed that both TOEFLs correlated highly with the MTELP/LCT, but the CBT ( r=.742) was higher than the PBT (r=.658). Since CAT was incorporated unevenly throughout the CBT and not at all on the PBT, multiple regression analysis of the two different versions' subtests was performed. This analysis revealed that the Listening Comprehension subtest of the CBT (beta=|.464|), which fully incorporated CAT, was responsible for most of the test's correlation with the MTELP/LCT, while Structure and Written Expression (beta=|.251|), which only partially employed CAT, and Reading Comprehension (beta=|.154|) which did not use it at all, had weaker impact on the correlation.; The main conclusion of this dissertation is that the effect of the computerization of the TOEFL on incoming international students at UM is reflected in statistical terms by increased measurement precision resulting mainly from the use of CAT in the Listening Comprehension subtest.