A Preliminary Study On The Competence Structure Of The Computer-based CET-4

With the advent of computer technology in language testing, the computer-based testing has been increasingly utilized with the traditional paper-based testing. The computer-based College English Test-4 (CB CET-4) is a computer-delivered test of students'English proficiency administered by the National College English Testing Committee. However, the validity of the new trial CB CET-4 has really become a great concern among researchers on language testing. For example, does the new test measure what it intends to measure, or is the new test in line with the test takers'computer proficiency. Therefore, more empirical researches are in need to gather validity evidence for this new test. As part of the validation study of CB CET-4, this paper undertakes to analyze the competence structure of CB CET-4.One crucial purpose of this paper is to investigate the competence structure of the CB CET-4 with a particular focus on two issues. The first is to examine to what extent the results of the computer-based CET-4 were comparable to those of the paper-based version. The second is to investigate which model fits the competence structure of the CB CET-4. A series of methods including statistical analysis, Pearson correlation, exploratory factor analysis and structural equation modeling were conducted for a field test completed by 448 participants in June, 2009. The results of data analyses were as follows:(1) Test scores of the paper-based test and the computer-based test were highly correlated. The paper-based test won in balancing the difficulty level of different subtests; while the computer-based version had a better internal structure and did a better job in differentiating test takers'English language proficiency.(2) As for the competence structure of the CB CET-4, among the four competitive models (one factor model, two-correlated factor model, three-correlated factor model and the higher-order factor model), a higher-order factor model, with a higher-order general factor (EFL ability) and three first-order factors for listening, listening-based integrated tasks and reading, was identified as the most fitting model.(3) The newly-introduced integrated tasks were discussed in particular. These tasks turned out to be closely related to the integrated ability factor, but had low correlations with the listening ability. So the author concluded that the integrated tasks were a measurement of integrated skills.In the conclusion chapter of this paper, the author also gave some suggestions for future development of the trial CB CET-4 in test format design.