The purpose of this study was to assess the effects of schema-broadening instruction (SBI), with and without explicit instruction in strategies for tackling the complexities involved in real-life (RL) math problems, on the math problem solving of third-grade students. Teachers (n = 30) were assigned randomly to 3 16-week conditions: control, SBI, and SBI-RL. Students in their classes (n = 455) were pre- and posttested on problem-solving measures at increasing transfer distances from the problems used for teaching problem solutions, with far transfer mirroring real-life problems. For the most part, SBI students, with and without RL, improved statistically significantly better than control students, with large effect sizes, but comparably to each other. By contrast, on the least structured real-life question, paralleling the ambiguities of real-life problem solving most closely, reliable and large effects favored the SBI-RL students over SBI and control students, suggesting RL's potential.