Cognitive diagnostic analyses have been advocated as methods that allow an assessment to function as a formative assessment to inform instruction. To use this approach, it is necessary to first identify the skills required for each item in the test, known as a Q-matrix. However, because the construct being tested and the underlying cognitive processes associated with it are usually not fully understood, establishing a Q-matrix, especially for an existing test, is a challenging task. This study reports the process of constructing and validating a Q-matrix for the reading comprehension section of the Michigan English Language Assessment Battery (MELAB). An initial Q-matrix was first generated based on evidence gathered from related literature, students' think-aloud protocols, and expert ratings. This initial Q-matrix was then validated empirically by applying the fusion model to a large MELAB data set. A well-supported Q-matrix was produced for potential future diagnostic applications. (Contains 10 tables and 4 figures.)