The purposes of this study were (1) to compare the differences of metacognition and mathematic performance between boys and girls. (2) to investigate the relationship between metacognition and mathematic performance. (3) to examine the differences of mathematic performance and problem solving ability between the high metacognition and low metacognition. The sample for the study included two hundred elementary school students. They were administered mathematic performance tests and metacognitive scales. The data was analyzed by t-test, Pearson product-monent correlation, canonical correlation, and Manova analysis. The major findings of this study were (1) there were no significant differences of metacognition and mathematic performance between the boys and girls. (2) there were significant correlations between metacognition and mathematic performance. (3) the higher the metacognition was, the higher the mathemtic performance was. All the results revealed that metacognitive knowledge and metacognitive process were important for mathematic performance.