Some mathematics educators have taken the position that metacognition- knowledge and beliefs about cognitive activity as well as awareness and control of that knowledge-is critical to mathematical performance and that research in mathematical problem solving would benefit from exploration of this phenomenon. This project is designed to (1) investigate the relationship between inservice middle school teachers' metacognitive activity and mathematical knowledge and their problem- solving ability, and (2) improve teachers' problem-solving ability through a problem-solving institute focusing on metacognitive knowledge and experience in mathematical problem solving. Fifteen middle school teachers enrolled in an institute on problem solving and thinking will be the subjects in the study. Subjects will be videotaped solving unfamiliar problems before small groups of their own students and will be tested for problem-solving performance before and after the institute. Techniques for recording and analyzing metacognitive activity are grounded in the work of Schoenfeld (1983), and problem-solving performance will be evaluated using a process/ product scale developed by Lucas et al.(1979) and Schultz (1985).