To investigate whether child × instruction effects contribute to mathematics gains in first grade, we conducted a cluster randomized controlled trial with 454 (205 treatment; 249 control) students in 28 classrooms. Classrooms were randomly assigned to implement Math Peer Assisted Learning Strategies (Math PALS) to supplement Saxon Math or Saxon Math alone. Teachers received professional development on either Math PALS (treatment) or individualized reading instruction (control). Intent-to-treat and treatment-on-the-treated results revealed no significant difference between the Math PALS and control groups in student mathematics gains. However, Math PALS effectiveness varied with initial math skills. Based on modeled results of the continuous fall math score, students who scored at the 75th percentile of the sample demonstrated significantly stronger mathematics outcomes compared with the treated control group; there was no significant treatment effect for students at the 50th percentile. For students with lower initial math skills (modeled at the 25th percentile of the sample), the effect of the treatment was negative. Students with lower initial mathematics skills made greater gains in the control group (Saxon Math alone) versus the Math PALS group.