Many students identified with a specific learning disability (SLD) experience difficulty with mathematical problem solving. Schema-based instruction (SBI) is identified as a promising practice for students with a SLD. The current projects aimed to tests the efficacy of SBI implemented by a teacher under routine conditions. This extends prior literature by (a) using a teacher as the implementer, (b) allowing flexibility in the intervention protocol, (c) condensing the duration of intervention sessions, and (d) providing instruction in small group settings. In addition, we measured student problem solving performance on word problems requiring two-steps. We used a multiple-probe design across three groups of fifth-grade participants (n = 7) receiving supplemental instruction in a resource room setting. Results indicated a functional relation between SBI and problem-solving performance on one-step word problems. However, the magnitude of change and consistency in performance varied across participants. Student performance on two-step word problems embedded within one structure and combined structures was low. The NAP, Tau, LRRi, and BC-SMD effect sizes were used to quantify effects. Implications are discussed in regard to systematic replication and conditions that may impact fidelity.