This article provides an introduction to Structural Equation Modeling (SEM) for language testing research. SEM is a way of
representing
hypothesized relationships among a set of substantively meaningful variables. It is also a multivariate analytic procedure for representing and testing (1) hypothesized inter-relationships between observed and latent variables and (2) hypothesized inter-relationships among latent variables. The procedures for testing the hypothesis of linkages between observed variables and their underlying latent variables are referred to as the measurement model and those for testing the hypotheses of linkages among latent variables are referred to as the structural model. The analysis of both the measurement and structural models together is called the full latent variable model. Therefore, SEM involves both a measurement model and/or a structural model. This paper discusses the commonly used steps in SEM, advantages of SEM over other multivariate procedures, statistical assumptions to be met, and some directions for future SEM applications in language testing research.