This article presents a brief history of the National Science Foundation's involvement in education evaluation, reviews the program evaluations that NSF's Division of Research, Evaluation, and Communication (REC) has conducted in recent years, and describes future directions. The evaluations have reflected the field-driven nature of NSF's programs, which are guided primarily by peer review. Three key issues have underlined NSF's policy regarding education evaluation: (1) the question must drive the methodology, with each evaluation expected to adhere to standards of evidence relevant to the approach chosen to answer the question; (2) there is a shortage of well-qualified evaluators for STEM (science, technology, engineering, and mathematics) education projects and programs; and (3) there is a serious lack of instruments of demonstrated validity and reliability to measure important outcomes of STEM education interventions, including teacher knowledge and skills, classroom practice, and student conceptual understanding in mathematics and science. (Contains 1 figure.)