We evaluated the psychometric properties of the Classroom Practices Survey--Revised (CPS-R) when used with students achieving at low, average, and high levels. A total of 739 teachers completed CPS-R for students in their classrooms. Results showed improvement in the reliability of CPS-R across all achievement levels when compared with its previous version. Internal consistency estimates for the four factors were higher for the high-achieving students ([alpha] = 0.84-0.94) compared with estimates for students who achieve at average ([alpha] = 0.83-0.92) and low ([alpha] = 0.81-0.90) levels. Model fit of the data was in the acceptable range across all achievement levels. However, model fit indices for the high-achieving group were slightly better than for the average- and low-achieving groups. Results support the practical value of CPS-R as a tool to assess teachers' use of differentiation strategies.