Abstract
Although L2 listening construct has been hypothesized to consist of the ability to understand literal messages and the ability to make inferences, its structure has not beenfully investigated empirically. One exception is Shimada (2000), which has explored the psychometric dimensionality of L2 listening ability and found that L2 listening ability is psychometrically more bidimensional. However, his study seems to suffer from a flaw in its methods of dimensionality analysis. The present research partially replicates Shimada and reexamines the psychometric dimensionality of L2 listening ability. The CELT was taken by 171 Japanese EFL college students and analyzed using item-level confirmatory factor analysis and testlet-based confirmatory factor analysis, which were used to overcome the problem with the analyses and differed from Shimada's methods. The results showed that L2 listening ability was psychometrically more bidimensional, which was congruent with Shimada. The results suggested that (a) when we want to measure the whole range of listening ability, two types of items should be put into a test, and that (b) it may be important to see whether data is psychometrically more unidimensional or bidimensional before using item response theory. The fact that the same conclusion as Shimada was reached using different methods reconfirms that L2 listening ability is psychometrically more bidimensional at least among Japanese EFL students.
