The majority of research on the efficacy of ClassWide Peer Tutoring (CWPT) is based on research with urban elementary students (Rohrbeck, Ginsberg-Block, Fantuzzo, & Miller, 2003), with much less research in middle schools. This study investigated CWPT with 975 middle school students in 52 classrooms, grades 6 through 8, over a three-year period. A mixed design combining features of both group (interrupted time-series) and single-subject reversal designs was used to evaluate the effects of traditional teacher-led instruction vs. CWPT. Results favored CWPT with effect sizes, based on weekly quizzes, indicating moderate to large effects overall (M = 1.11) but with some range across classrooms and content. Implications for future research and practice are discussed. (Contains 2 figures and 5 tables.)