While Grosse and Voght (1991) set out a well-considered overview of LSP and identified areas in need of development, they limited their observations on the topic of assessment to a short section devoted to what they called the "proficiency movement." While it is true that they really did not have a lot to report on at the time they wrote their review, and that this current review highlights much interesting work done in the area in the intervening years, there remains a considerable emphasis on practice rather than on theory, with published papers attempting to identify solutions to given assessment "problems," rather than attempting to build cohesive assessment "theories." The primary focus of this contribution is on the latter aspect of language assessment. The article first offers a historical overview of the area over the past twenty years, which moves from a brief discussion of the issues highlighted by Grosse and Voght to the theoretical issues that have emerged, and finishing with a critical review of research on issues around assessment in three specific domains (immigration and citizenship, or work and the professions). The article then highlights current needs and priorities, focusing on issues of test usage and introducing the concept of test localization before presenting the core of the argument: a theory of LSP assessment validation. It concludes with a preliminary attempt to exemplify how a theory-driven research agenda can inform future research, and ultimately, practice in the area of LSP assessment over the coming decades. (Contains 3 figures and 12 online resources.)