earticle

영어

This study examined the comparability of three L2 oral test tasks (topic discussion, information gap, and semi-direct speaking) in the measurement of the four oral traits (pronunciation, fluency, grammar, and vocabulary). In examining the comparability, a CFA approach to a correlated traits/correlated methods design was employed, and evidence of convergent and discriminant validity was tested in the measurement of the traits by the methods. Through the model comparison approach (Byrne, 2006; Wideman, 1985), the hypothesized model was compared against a series of more restrictive models using χ2 difference tests, and the information from the path coefficients was examined to check the equivalence further. The findings revealed that the method effect in the measurement model was so strong that the trait was not fully reflected in the measurement process. Also, the effect of method on trait was not consistent in that topic discussion produced most trait-related information among the test tasks. Moreover, methods were found discriminant although traits were not. Thus, the three methods cannot be considered comparable in measuring the L2 oral traits, and such incomparability was particularly evident between the two methods of group oral and semi-direct speaking.