Standardization has become a ubiquitous feature in the field of education both through federal initiatives, such as the establishment of “best practices” distilled from randomized controlled trials (RCTs), and through private businesses and non-profits, such as Facebook and the Gates Foundation, which create and disseminate prefabricated curricula and standardized software programs. Standardization requires, as a precondition, the generalization of research findings from a smaller subset of teachers, students, or schools, to the field of education in its entirety. This dissertation investigates whether generalization of this sort is possible or desirable. After explaining why current critiques of educational generalization are insufficient, the author argues that, though generalizing from RCTs is ontologically precluded, generalization of a different sort is both possible and desirable. The author employs Martin Heidegger’s ontological analysis of language to argue for a weak form of generalizability that avoids the extremes of RCT-based best practices while allowing teaching to be discussed across spatial and temporal locations.