Language testing has witnessed three major trends in the 1990s: theoretical, methodological, and analytical. Theoretically, emphasis has been placed on the further understanding of the construct of language proficiency. Methodologically, there has been an outburst of interest in language performance testing and the promotion of the professional standards of test development and use. Analytically, emphasis has been placed on the implementation of Item Response Theory (IRT), G-theory, and the understanding of the multiple sources of variance in test performance. After a review of these trends, the current paper presents a complete picture of language assessment from an Archimedean point. It argues that language assessment, seen from such a point, has four intertwined but self-informed pillars: construct issues, psychometrics, edumetrics, and construct-irrelevant factors.