Metacognitive monitoring and regulation play an essential role in mathematical problem solving. Therefore, it is important for researchers and practitioners to assess students' metacognition. One proven valid, but time consuming, method to assess metacognition is by using think-aloud protocols. Although valuable, practical drawbacks of this method necessitate a search for more convenient measurement instruments. Less valid methods that are easy to use are self-report questionnaires on metacognitive activities. In an empirical study in grade five (n = 39), the accuracy of students' performance judgments and problem visualizations are combined into a new instrument for the assessment of metacognition in word problem solving. The instrument was administered to groups of students. The predictive validity of this instrument in problem solving is compared to a well-known think-aloud measure and a self-report questionnaire. The results first indicate that the questionnaire has no relationship with word problem solving performance, nor the other two instruments. Further analyses show that the new instrument does overlap with the think-aloud measure and both predict problem solving. But, both instruments also have their own unique contribution to predicting word problem solving. The results are discussed and recommendations are made to further complete the practical measurement instrument. (Contains 1 figure and 3 tables.)