Abstract: This article contends that the necessary links between constructs and test scores/decisions in language assessment must be established through principled design procedures that align three models: (1) a theoretical construct model; (2) a test development model; and (3) a psychometric scoring model. The theoretical construct model articulates the cognitive tasks, mechanisms, and ability constructs underlying language proficiency. The test development model focuses on developing formal mechanisms for structuring the design and development of the test problems (items and tasks) and for assembling the test as a whole. The psychometric scoring final model involves the use of psychometric modeling techniques capable of handling complex, multidimensional structures underlying the responses to a language proficiency assessment. The authors promote the need for modeling the complexity inherent in the process of defining constructs, designing and building problems and tests, and finally, scaling and scoring the instruments that assess language proficiency.