A key tool in multitiered systems of support is the use of curriculum-based measures to predict which students are at risk for academic failure. However, there are few studies that examine which measures are most suitable for students in middle school. The authors examine the reliability of predicting outcomes on state assessments for 3 commonly used curriculum-based measures at the middle school level. Data were collected from a middle school in the Midwest that regularly administers 3 different curriculum-based measures. Reading Curriculum-Based Measure (R-CBM), Maze, and Multiple-Choice Reading Comprehension (MCRC) were given to students in Grades 7 (n = 238) and 8 (n = 256). Logistic regression was used to examine each measure in predicting outcomes on the Michigan Education Assessment Program Reading assessment. Results indicated that MCRC more accurately predicted outcomes than R-CBM or maze (Grade 7 e[superscript ß] = 1.75, Grade 8 e[superscript ß] = 1.68). Limitations and recommendations for future research are discussed.