Students are increasingly being asked to demonstrate their conceptual knowledge through enacting the practices that scientists and engineers engage in. The Next Generation Science Standards–adopted by 19 states and the District of Columbia–have accelerated this shift. Assessments that allow students to demonstrate their knowledge-in-use are critical because they (a) signal to teachers and students what proficiency “looks like” and (b) can support instruction that integrates science and engineering practices with content knowledge. In this talk I focus on two aspects of validating classroom assessments: design and modeling. I will describe and exemplify a principled assessment design process–evidence-centered design–that was used to develop assessments for middle school life and physical science. I will report on an empirical study in which we used a measurement model to analyze the performance of assessment tasks. Together, these design and modeling efforts support our validity argument for the use of these multi-dimensional assessment tasks.