Various stakeholders, such as policymakers and educators, require diagnostic feedback and actionable test results. One particular context in which fine-grained test results are of utmost importance is English language proficiency (ELP) assessments as they are used for critical decisions for students. Diagnostic classification models (DCMs) afford finer levels of feedback to improve learning outcomes. This study is part of a research project which implemented DCMs to the reading domain of a K-12 ELP assessment for grades 6-8 to evaluate its feasibility and utility. The research project is U.S.-based. However, the investigation is of relevance to the language testing community broadly. The paper discusses how to address second language reading construct within DCMs. Specifically, it details the identification of attributes underlying the construct and the development and selection of alternative Q-matrices.