This study proposed various models with different constructs of the Foreign Language Classroom Anxiety Scale (FLCAS) developed by Horwitz et al. (1986) and tested the competitive models with a view to finding the constructs that best account for the FLCAS. The FLCAS was administered to 918 EFL university students in Korea, and the data were analyzed by performing confirmatory factor analysis (CFA). The resu lts of data analyses indicated that even though the fit indices for all the hypothesized models in general did not meet the cutoff points of acceptabi lity, the model containing four constructs fit the data better than the models containing one, two, or three constructs of the FLCAS. Implications of these findings followed by futu re research areas were provided to deepen the insight into foreign language anxiety in the classroom in general and the FLCAS in particular.