Recent large-scale efforts to improve mathematics education have focused on changing teachers’ instructional practices to place less emphasis on the acquisition of discrete skills and factual knowledge, and greater weight on conceptual understanding, inquiry, and application and communication of mathematical or scientific ideas. Research has shown that this approach, commonly referred to as reform-oriented instruction, can have positive but small effects on student achievement. The weak relationships between reform pedagogy and achievement may be partially attributable to several factors, including inadequate measurement of reform instruction, a brief timeframe in which reform teaching is examined, and the use of achievement measures (such as multiple-choice tests) that may not be well-aligned with reform practices or curricula. This study addresses those limitations by using innovative vignette-based methods for measuring instructional practices, exploring relationships longitudinally over a three-year period, and including open-ended measures. Analyses suggested that some aspects of reform teaching show positive relationships with mathematics achievement, but the effects are quite small, may take several years before the relationships are manifested, and may be evident only on certain test formats or grade levels. The results help inform the debate regarding the relative importance of teachers as a source of variability in student achievement, and provide direction for future research.