Students with the most significant cognitive disabilities (SCD) are the 1% of the total student population who have a disability or multiple disabilities that significantly impact intellectual functioning and adaptive behaviors and who require individualized instruction and substantial supports. Historically, these students have received little instruction in science and the science assessments they have participated in have not included age-appropriate science content. Guided by a theory of action for a new assessment system, an eight-state consortium developed multidimensional alternate content standards and alternate assessments in science for students in three grade bands (3-5, 6-8, 9-12) that are linked to the Next Generation Science Standards (NGSS Lead States, 2013) and A Framework for K-12 Science Education (Framework; National Research Council, 2012). The great variability within the population of students with SCD necessitates variability in the assessment content, which creates inherent challenges in establishing technical quality. To address this issue, a primary feature of this assessment system is the use of hypothetical cognitive models to provide a structure for variability in assessed content. System features and subsequent validity studies were guided by a theory of action that explains how the proposed claims about score interpretation and use depend on specific assumptions about the assessment, as well as precursors to the assessment. This paper describes evidence for the main claim that test scores represent what students know and can do. We present validity evidence for the assumptions about the assessment and its precursors, related to this main claim. The assessment was administered to over 21,000 students in eight states in 2015-2016. We present selected evidence from system components, procedural evidence, and validity studies. We evaluate the validity argument and demonstrate how it supports the claim about score interpretation and use.